review fixes

Give each scan its own DNS resolvers

DESIGN.md

@@ -36,7 +36,7 @@ It also specifies a `WorkflowLoader` interface that is used during workflow load
 
 ```go
 type WorkflowLoader interface {
-	GetTemplatePathsByTags(tags []string) []string
+	GetTemplatePathsByTags(tags []string, noValidate bool) []string
 	GetTemplatePaths(templatesList []string, noValidate bool) []string
 }
 ```

README.md

@@ -145,12 +145,12 @@ TARGET:
    -eh, -exclude-hosts string[]  hosts to exclude to scan from the input list (ip, cidr, hostname)
    -resume string                resume scan from and save to specified file (clustering will be disabled)
    -sa, -scan-all-ips            scan all the IP's associated with dns record
-   -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
+   -iv, -ip-version string[]     IP version to scan of hostname (4,6); default uses dialer fallback, -sa scans all addresses
    -dun, -disable-url-normalization  keep targets exactly as supplied instead of collapsing urls that address the same resource
    -mup, -max-urls-per-pattern int   scan at most this many urls sharing one structural shape, e.g. /user/1 and /user/2 (0 to scan all)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -177,7 +177,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -221,6 +221,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file
@@ -326,6 +327,7 @@ OPTIMIZATIONS:
    -nh, -no-httpx                   disable httpx probing for non-url input
    -preflight-portscan              run preflight resolve + TCP portscan and filter targets before scanning (disabled by default)
    -no-stdin                        disable stdin processing
+   -ac, -auto-calibrate             drop HTTP matches on a host's catch-all page (answers missing paths with 2xx/3xx)
 
 HEADLESS:
    -headless                        enable templates that require headless browser support (root user on Linux will disable sandbox)

README_CN.md

@@ -148,7 +148,7 @@ TARGET:
    -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -175,7 +175,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -219,6 +219,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

README_ES.md

@@ -148,7 +148,7 @@ TARGET:
    -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -175,7 +175,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -219,6 +219,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

README_FA.md

@@ -165,7 +165,7 @@ TARGET:
    -mup, -max-urls-per-pattern int   scan at most this many urls sharing one structural shape, e.g. /user/1 and /user/2 (0 to scan all)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -192,7 +192,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -236,6 +236,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

README_ID.md

@@ -148,7 +148,7 @@ TARGET:
    -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -175,7 +175,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -219,6 +219,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

README_JP.md

@@ -148,7 +148,7 @@ TARGET:
    -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -175,7 +175,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -219,6 +219,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

README_KR.md

@@ -148,7 +148,7 @@ TARGET:
    -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -175,7 +175,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -219,6 +219,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

README_PT-BR.md

@@ -148,7 +148,7 @@ TARGET:
    -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -175,7 +175,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -219,6 +219,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

README_TR.md

@@ -148,7 +148,7 @@ TARGET:
    -iv, -ip-version string[]     IP version to scan of hostname (4,6) - (default 4)
 
 TARGET-FORMAT:
-   -im, -input-mode string        mode of input file (list, burp, jsonl, yaml, openapi, swagger, http) (default "list")
+   -im, -input-mode string        mode of input file (list, burp, har, jsonl, yaml, openapi, swagger, http) (default "list")
    -ro, -required-only            use only required fields in input format when generating requests
    -sfv, -skip-format-validation  skip format validation (like missing vars) when parsing input file
    -vtt, -vars-text-templating    enable text templating for vars in input file (only for yaml input mode)
@@ -175,7 +175,7 @@ TEMPLATES:
    -egm, -enable-global-matchers          enable loading global matchers templates
    -file                                  enable loading file templates
    -llm                                   enable llm matchers and extractors (semantic matching)
-   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter) (default "openai")
+   -llm-provider string                   llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev) (default "openai")
    -llm-base-url string                   openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)
    -llm-model string                      model used for semantic matching
    -llm-timeout int                       time in seconds to wait for a single llm call (default 30)
@@ -219,6 +219,7 @@ OUTPUT:
    -jle, -jsonl-export string    file to export results in JSONL(ine) format
    -pe, -pdf-export string       file to export results in PDF format
    -rd, -redact string[]         redact given list of keys from query parameter, request header and body
+   -nrd, -no-redact              disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted
 
 CONFIGURATIONS:
    -config string                        path to the nuclei configuration file

SECURITY_CONTEXT.md

@@ -1,255 +1,120 @@
-<!-- Generated by securitycontext.dev. Source HEAD: 3e1e0e81f091c7817bbf7af64bb8eec6e8c2bac1. -->
+<!-- Generated by securitycontext.dev. Source HEAD: df57ae6d2955864fd92a1bd2565fe9e4e9a7592b. -->
 # Security Context: projectdiscovery/nuclei @ dev
 
 > Reference context for editing projectdiscovery/nuclei. Before changing any component below, **preserve the listed guards and do not reintroduce these fixed issues**. Each fingerprint is a regression check, and each shared attack surface is a guard that must keep holding on every path.
 
-_Generated 2026-09-01 16:22 UTC by secc_
-
-_Analyzed commit: 3e1e0e81f091c7817bbf7af64bb8eec6e8c2bac1_
-
-**Remediation coverage: 56.7%**
-_17 of 30 historical fixes have no remaining unguarded variant; 368 open remediation leads to verify. Measures fix consistency, not a vulnerability verdict._
-
-- Commits analyzed: **4814**
-- Commits flagged for review: **235**
-- Confirmed historical security fixes: **32**
-- Known published CVEs: **0**
-
-## Summary
-
-Nuclei's historical security posture is dominated by sandbox escapes and code execution via unsigned or unverified templates, with recurring SSRF, path traversal, and auth bypass issues across JS libraries, loaders, and workflow paths. The most critical theme is inconsistent enforcement of signature verification and local-file-access (-lfa) guards across direct, workflow, DAST, and JS-sandbox entry points, enabling variant attacks. A second theme is network policy (host allowlist) gaps in LDAP dialing and proxy handling. The security context also includes dependency upgrades (Alpine, go-git) and output file permission hardening, but the highest-value focus is on template trust boundaries and sandbox enforcement parity.
+_Analyzed [df57ae6d29][c-df57ae6d29] on `dev` · 5174 commits, 548 security-relevant examined · generated 2026-11-02_
 
 ## Recurring Security Patterns
 
-**15.6% of classified fixes repeat a previously fixed pattern** (5 recurring, 27 first-seen). A recurring pattern means this repository fixed a similar issue before — not that the code is currently vulnerable.
-
-- **Code Execution:** [1c440e755a](https://github.com/projectdiscovery/nuclei/commit/1c440e755a97471c56fb0276ee8a8c4132974645) → [9de96e4dda](https://github.com/projectdiscovery/nuclei/commit/9de96e4dda5a03da963b9ae6582f03ea55791a76) — matched on `class:code execution`, `component:pkg/catalog/loader/loader.go`
-- **Code Execution via Unsigned Template:** [fa56800fcc](https://github.com/projectdiscovery/nuclei/commit/fa56800fcca3c8c8fe8e8dec28584210ca1527de) → [fac0b7196d](https://github.com/projectdiscovery/nuclei/commit/fac0b7196d2fbca56c8af8ff8e1899c111041a37) — matched on `class:code execution via unsigned template`, `CWE-347`, `sink:loadtemplates`
-- **SSRF:** [a167e6c57b](https://github.com/projectdiscovery/nuclei/commit/a167e6c57b80f5dc7be231231b62d7c594ee2770) → [9def9af7e3](https://github.com/projectdiscovery/nuclei/commit/9def9af7e3f7ebcf5cd561c37b623b9999ee7572) — matched on `class:ssrf`, `CWE-918`, `component:pkg/js/libs/ldap/ldap.go`
-- **Dependency Vulnerability:** [d93b4a01df](https://github.com/projectdiscovery/nuclei/commit/d93b4a01df9c218a36dd2fccfb1c2c1f96b181fe) → [fcd5c6b111](https://github.com/projectdiscovery/nuclei/commit/fcd5c6b111ddcb746525f88b7d0bf4b4abbc2529) — matched on `class:dependency vulnerability`, `CWE-1104`, `component:dockerfile`
-- **Sandbox Escape:** [96049a32fa](https://github.com/projectdiscovery/nuclei/commit/96049a32fa70aef0483d4180d386a57b3c2d973a) → [cf2a98b03d](https://github.com/projectdiscovery/nuclei/commit/cf2a98b03d76327c20fb0183b7f1c849eb692c91) — matched on `class:sandbox escape`, `CWE-284`, `component:v2/internal/runner/processor.go`, `sink:stdlib allmodulenames`, `sink:stdlib getmodulemap`
+**7.1% of classified fixes repeat a previously fixed pattern** (1 recurring, 13 first-seen). A recurring pattern means this repository fixed a similar issue before — not that the code is currently vulnerable.
 
 ## Top Risks
 
-### 1. Sandbox Escape: `pkg/js/compiler/pool.go, v2/pkg/protocols/javascript/js.go, v2/pkg/protocols/headless/engine/engine.go, v2/internal/runner/processor.go` (high)
+Components with the strongest concentration of past fixes for one class.
 
-Multiple sandbox escapes across JS require(), eval/ReadFile, headless browser flags, and tengo imports, indicating weak default sandbox boundaries that require consistent -lfa and safe-module enforcement.
-
-Evidence: 6f2ade6a9b427c284c15a43445f9c7f055e60e5d, c35162c8ef061a96dd74b54c8f022c1a41cf5833, df55f7a2eb591a7949b5c66a9c54db54aa28f450, cf2a98b03d76327c20fb0183b7f1c849eb692c91, 96049a32fa70aef0483d4180d386a57b3c2d973a
-
-### 2. Code Execution via Unsigned Template: `pkg/catalog/loader/loader.go, pkg/templates/workflows.go, pkg/templates/templates.go` (critical)
-
-Unsigned templates can execute code through direct loading, workflows, and DAST branches; signature verification was bypassed via cache poisoning and missing guards on specific paths.
-
-Evidence: 9de96e4dda5a03da963b9ae6582f03ea55791a76, 1c440e755a97471c56fb0276ee8a8c4132974645, fac0b7196d2fbca56c8af8ff8e1899c111041a37, e86f38299765b82ad724fdb701557e0eaff3884d, fa56800fcca3c8c8fe8e8dec28584210ca1527de
-
-### 3. Auth Bypass: `pkg/templates/compile.go, pkg/templates/signer/tmpl_signer.go, pkg/catalog/loader/loader.go, pkg/protocols/http/request.go` (high)
-
-Signature verification can be bypassed via cached metadata, improper signature extraction, or missing checks on specific load paths, and cookie jar bypass via disabled cookie handling.
-
-Evidence: 4534e9cb3034ebe5cc7b0cecabe399b6fb9ce0e4, 0da993afe6d41b4b1b814e8fad23a2acba13c60a, 9bd4db3f74aedd067af568df4f9198a8a090f68b, a976ca8d3f6a18c9ca746c8ce73f28fdcd518642
-
-### 4. SSRF: `pkg/js/libs/ldap/ldap.go, v2/internal/runner/proxy.go` (high)
-
-SSRF via LDAP dialing to internal hosts and via proxy configuration; host allowlist checks were missing or applied to the wrong target (domain vs controller).
-
-Evidence: 9def9af7e3f7ebcf5cd561c37b623b9999ee7572, a167e6c57b80f5dc7be231231b62d7c594ee2770, 0e46d3e0410f89e4af41099f5cd38fa7f54ae54f
-
-### 5. Path Traversal: `pkg/catalog/config/nucleiconfig.go, v2/internal/runner/update.go, v2/pkg/protocols/common/protocolstate/headless.go` (high)
-
-Path traversal via lexical prefix checks, zip-slip in template extraction, and file:// scheme in headless mode; canonicalization and containment checks are required.
-
-Evidence: 20cbf0bd73893c1270582c9778b6b26a5cb11206, f9c214a66f37a00fc29fe64232b690e6285ec59c, d3928e080d819f25306bfbaecc399f03516db5aa
-
-### 6. Code Execution: `pkg/catalog/loader/loader.go, pkg/templates/workflows.go` (high)
-
-Code execution via cache poisoning and unverified code templates in workflows; requires -code flag and verified templates.
-
-Evidence: 9de96e4dda5a03da963b9ae6582f03ea55791a76, e86f38299765b82ad724fdb701557e0eaff3884d
-
-## Shared Attack Surfaces
-
-Capabilities guarded by more than one past fix on different code paths. A guard enforced on one entry point is often missing on a sibling; verify each alternate path before assuming it's covered.
-
-### 1. Template signature verification and unsigned-template opt-in enforcement across all load paths
-
-- Guard that should hold everywhere: signature verification (template.Verified, Verify), DisableUnsignedTemplates flag, -code flag for code templates, cache invalidation on verification status/fingerprint/digest
-- Entry points: `pkg/catalog/loader/loader.go (LoadTemplatesWithTags, DAST branch)`, `pkg/templates/workflows.go (loadTemplates, workflowExecutor)`, `pkg/templates/templates.go (loadTemplates)`, `pkg/templates/compile.go (template.Verified)`, `pkg/templates/signer/tmpl_signer.go (Verify)`
-- Check: Inspect the DAST and workflow sub-loaders: they must enforce the same DisableUnsignedTemplates and -code checks as the direct loader, and must not rely on cached verification results without re-checking signatures.
-- Evidence: 9de96e4dda5a03da963b9ae6582f03ea55791a76, 1c440e755a97471c56fb0276ee8a8c4132974645, fac0b7196d2fbca56c8af8ff8e1899c111041a37, e86f38299765b82ad724fdb701557e0eaff3884d, fa56800fcca3c8c8fe8e8dec28584210ca1527de, 4534e9cb3034ebe5cc7b0cecabe399b6fb9ce0e4, 0da993afe6d41b4b1b814e8fad23a2acba13c60a, 9bd4db3f74aedd067af568df4f9198a8a090f68b
-
-### 2. Local file access (read/write) outside the sandbox, gated by -lfa flag
-
-- Guard that should hold everywhere: -lfa flag enforcement, path canonicalization and containment (sandbox directory), file:// scheme blocking, allowAllFiles stripping, require() registry rebuild
-- Entry points: `pkg/js/libs/krbforge/krbforge.go (CreateTicket)`, `pkg/js/libs/mysql/mysql_private.go (sql.Open)`, `pkg/js/compiler/pool.go (require)`, `v2/pkg/protocols/javascript/js.go (ReadFile, eval)`, `v2/pkg/protocols/common/protocolstate/headless.go (file:)`, `v2/pkg/protocols/common/generators/load.go (loadPayloadsFromFile)`
-- Check: Verify that every JS library and protocol helper (krbforge, mysql, require, headless file://, payload loaders) enforces the -lfa flag and path containment; a sibling path that skips the -lfa check is a sandbox escape.
-- Evidence: 3a8b65cb45f121fcdf418f60e5177ed9f250c122, 9611926a9a1ca99e8ed554d1b26fbeff60e06368, 6f2ade6a9b427c284c15a43445f9c7f055e60e5d, c35162c8ef061a96dd74b54c8f022c1a41cf5833, d3928e080d819f25306bfbaecc399f03516db5aa, e7fb40a41308f6c41267e858d1781eb67f7e0572
-
-### 3. Network policy (host allowlist) enforcement before outbound connections
-
-- Guard that should hold everywhere: IsHostAllowed check on the actual target host (controller, not domain), proxy validation and connectivity check, ProxyInternal gate for HTTP_PROXY env
-- Entry points: `pkg/js/libs/ldap/ldap.go (Dial, newLdapSession)`, `v2/internal/runner/proxy.go (loadProxyServers, os.Setenv HTTP_PROXY)`
-- Check: Inspect other dialing paths (e.g., HTTP, headless) to ensure they apply the same host allowlist check on the resolved IP/host, not just a user-supplied domain, and that proxy settings cannot redirect traffic to internal hosts.
-- Evidence: 9def9af7e3f7ebcf5cd561c37b623b9999ee7572, a167e6c57b80f5dc7be231231b62d7c594ee2770, 0e46d3e0410f89e4af41099f5cd38fa7f54ae54f, 3f5c055aca1499642caf4f40e8cf55ad32bd4da9
-
-### 4. Path traversal in file extraction and template path trust boundaries
-
-- Guard that should hold everywhere: canonical path containment check (filepath.Rel, prefix), zip-slip validation against template directory, sandbox path prefix validation
-- Entry points: `pkg/catalog/config/nucleiconfig.go (IsCustomTemplate)`, `v2/internal/runner/update.go (os.OpenFile)`, `v2/pkg/protocols/common/generators/load.go (loadPayloadsFromFile)`
-- Check: Audit all file-write and file-read paths that accept template-controlled paths (e.g., template extraction, payload loading) to ensure they use canonical containment, not lexical prefix checks, and apply the same sandbox directory restriction.
-- Evidence: 20cbf0bd73893c1270582c9778b6b26a5cb11206, f9c214a66f37a00fc29fe64232b690e6285ec59c, e7fb40a41308f6c41267e858d1781eb67f7e0572
-
-## Historical Vulnerability Classes
-
-| Class | Count |
-|---|---|
-| Sandbox Escape | 7 |
-| Auth Bypass | 4 |
-| Code Execution | 3 |
-| Dependency Vulnerability | 3 |
-| Path Traversal | 3 |
-| SSRF | 3 |
-| Code Execution via Unsigned Template | 2 |
-| Arbitrary File Read | 1 |
-| Cryptographic Weakness | 1 |
-| Deserialization of Untrusted Data | 1 |
-| Information Disclosure | 1 |
-| Insecure File Permissions | 1 |
-| Race Condition | 1 |
-| Security Misconfiguration | 1 |
+| # | Class | Component | Severity | Fixes | Evidence |
+|---|---|---|---|---|---|
+| 1 | SSRF | `pkg/protocols/common/protocolstate/headless.go` | high | 2 | [d43c1f547f][c-d43c1f547f], [d9dabbb4a1][c-d9dabbb4a1] |
+| 2 | Arbitrary File Access | `v2/pkg/protocols/common/generators/load.go` | high | 0 (1 component gone, not counted) | [e5154d362a][c-e5154d362a] |
+| 3 | Arbitrary File Access | `pkg/js/libs/mysql/mysql.go` | high | 1 | [634cecbb28][c-634cecbb28] |
+| 4 | Code Execution | `pkg/fuzz/execute.go` | high | 1 | [ccbfb12bd0][c-ccbfb12bd0] |
+| 5 | Code Execution | `pkg/protocols/common/expressions/expressions.go` | high | 1 | [d221732016][c-d221732016] |
+| 6 | Information Disclosure | `internal/httpapi/apiendpoint.go` | high | 1 | [a20aa26024][c-a20aa26024] |
 
 ## Fixed-Issue Fingerprints
 
-These are real security fixes mined from git history. Use them as regression checks.
+Real security fixes mined from git history; use them as regression checks. `verified` fixes are linked to a published advisory, `inferred` fixes were confirmed from the diff evidence alone. Each is identified by its short commit SHA, which the variant-leads report uses to reference it back to this history.
 
-- **Auth Bypass** (high) [CWE-287]: Honors disable-cookie by preventing the shared cookie jar from being applied when cookie handling is disabled, closing a path where cookies could be reused across requests.
-  - sink: `SetCookieJar(input.CookieJar)`
-  - fix: Adds !request.DisableCookie check before applying shared cookie jar
-  - commit: https://github.com/projectdiscovery/nuclei/commit/a976ca8d3f6a18c9ca746c8ce73f28fdcd518642 (2026-08-25)
-- **Code Execution** (high) [CWE-347]: This commit fixes a cache poisoning vulnerability where unsigned templates could bypass signature verification and execute malicious code during compilation or request execution.
-  - sink: `template compilation and execution of JavaScript`
-  - fix: cache invalidation based on verification status, verifier fingerprint, and content digest
-  - commit: https://github.com/projectdiscovery/nuclei/commit/9de96e4dda5a03da963b9ae6582f03ea55791a76 (2026-08-04)
-- **Code Execution via Unsigned Template** (high) [CWE-347]: Enforces signature verification on JavaScript templates, preventing execution of unsigned templates that could lead to arbitrary code execution.
-  - sink: `loading unsigned javascript templates into execution paths`
-  - fix: Require signature verification for javascript templates in direct loading and workflows when signing is enforced
-  - commit: https://github.com/projectdiscovery/nuclei/commit/fac0b7196d2fbca56c8af8ff8e1899c111041a37 (2026-07-03)
-- **Sandbox Escape** (high) [CWE-22]: Enforces nuclei's local file access policy on krbforge output paths, preventing arbitrary file writes outside the sandbox.
-  - sink: `gpkrb.CreateTicket(cfg) with cfg.OutputFile`
-  - fix: Path normalization into sandbox and denial of outside paths unless -lfa
-  - commit: https://github.com/projectdiscovery/nuclei/commit/3a8b65cb45f121fcdf418f60e5177ed9f250c122 (2026-06-24)
-- **SSRF** (high) [CWE-918]: Adds network policy host allowlist checks before LDAP dialing, preventing SSRF to restricted internal hosts.
-  - sink: `dialers.Fastdialer.Dial`
-  - fix: host allowlist check via IsHostAllowed before dialing
-  - commit: https://github.com/projectdiscovery/nuclei/commit/9def9af7e3f7ebcf5cd561c37b623b9999ee7572 (2026-06-24)
-- **Arbitrary File Read** (high) [CWE-22]: Gates MySQL's allowAllFiles option behind the -lfa flag, preventing arbitrary local file reads via LOAD DATA LOCAL INFILE.
-  - sink: `sql.Open("mysql", dsn) with AllowAllFiles=true`
-  - fix: DSN sandbox that strips allowAllFiles unless -lfa is set
-  - commit: https://github.com/projectdiscovery/nuclei/commit/9611926a9a1ca99e8ed554d1b26fbeff60e06368 (2026-06-19)
-- **Code Execution** (critical) [CWE-94]: Moves the unsigned code-template guard ahead of the DAST branch, preventing unsigned code templates from being loaded and executed under DAST, which previously bypassed the signature check.
-  - sink: `Template loading and execution`
-  - fix: Hoisting unsigned code-template guard before DAST branch to ensure it applies on all load paths
-  - commit: https://github.com/projectdiscovery/nuclei/commit/1c440e755a97471c56fb0276ee8a8c4132974645 (2026-06-19)
-- **Path Traversal** (high) [CWE-22]: This commit replaces lexical prefix checks with canonical path containment checks across multiple filesystem trust boundaries, preventing path traversal where sibling paths were incorrectly treated as trusted.
-  - sink: `IsCustomTemplate`
-  - fix: path canonicalization and containment check
-  - commit: https://github.com/projectdiscovery/nuclei/commit/20cbf0bd73893c1270582c9778b6b26a5cb11206 (2026-06-04)
-- **Information Disclosure** (medium) [CWE-200]: This commit fixes a data leakage issue in workflows where sibling state could leak across concurrent executions, and race requests ignored dynamic values, by syncing context and cloning inputs.
-  - sink: `syncWorkflowInputToTemplateCtx`
-  - fix: context synchronization and input cloning
-  - commit: https://github.com/projectdiscovery/nuclei/commit/805bcfe76f3629dbe71021795d4ce865406c1555 (2026-05-08)
-- **Sandbox Escape** (high) [CWE-284]: Fixes sandbox escape where require() bypassed local file access restrictions and reused privileged module cache across executions.
-  - sink: `require() using default host filesystem loader`
-  - fix: Rebuild require registry per execution and enforce allow-local-file-access sandbox rules
-  - commit: https://github.com/projectdiscovery/nuclei/commit/6f2ade6a9b427c284c15a43445f9c7f055e60e5d (2026-04-10)
-- **Cryptographic Weakness** (medium) [CWE-338]: Replaces predictable math/rand PRNG with cryptographically secure crypto/rand in JS helper functions.
-  - sink: `rand.Intn(255) / rand.Int63()`
-  - fix: Replace predictable math/rand with crypto/rand
-  - commit: https://github.com/projectdiscovery/nuclei/commit/44ef9ce34ffcfdfca3a96626abc73d99dd0e83ab (2026-03-15)
-- **Auth Bypass** (high) [CWE-347]: Caching signature verification results without cryptographic re-check allows a metadata index compromise to bypass template signature verification.
-  - sink: `template.Verified, template.TemplateVerifier`
-  - fix: cached verification data reuse
-  - commit: https://github.com/projectdiscovery/nuclei/commit/4534e9cb3034ebe5cc7b0cecabe399b6fb9ce0e4 (2026-01-21)
-- **Race Condition** (medium) [CWE-362]: Fixes a race condition in workflow execution caused by shared context callbacks, which could lead to data corruption or crashes; the primary impact is reliability, not a direct security breach.
-  - sink: `shared context callbacks in workflow execution`
-  - fix: fix race condition by making workflow execution sequential
-  - commit: https://github.com/projectdiscovery/nuclei/commit/24d1f58682027e23cb21a299a170abd2bdf20dc2 (2025-12-05)
-- **Auth Bypass** (high) [CWE-347]: Fixes template signature verification by centralizing signature extraction and requiring the signature to be at the end of the template, preventing bypass.
-  - sink: `Verify`
-  - fix: signature extraction and verification
-  - commit: https://github.com/projectdiscovery/nuclei/commit/0da993afe6d41b4b1b814e8fad23a2acba13c60a (2024-08-19)
-- **Dependency Vulnerability** (medium) [CWE-1104]: Upgrades Alpine base image from 3.18.5 to 3.18.6 to fix known security vulnerabilities in bundled packages.
-  - sink: `FROM alpine:3.18.5`
-  - fix: Upgrade base image to patched version
-  - commit: https://github.com/projectdiscovery/nuclei/commit/fcd5c6b111ddcb746525f88b7d0bf4b4abbc2529 (2024-03-15)
-- **Dependency Vulnerability** (medium) [CWE-1104]: Upgrades Alpine base image from 3.18.2 to 3.18.5 to fix known security vulnerabilities in bundled packages.
-  - sink: `FROM alpine:3.18.2`
-  - fix: Upgrade base image to patched version
-  - commit: https://github.com/projectdiscovery/nuclei/commit/d93b4a01df9c218a36dd2fccfb1c2c1f96b181fe (2024-03-15)
-- **Dependency Vulnerability** (high) [CWE-1035]: Updates go-git to fix known vulnerabilities (CVE-2023-49569, CVE-2023-49568), which could be exploited when processing untrusted git data.
-  - sink: `go-git library usage`
-  - fix: dependency upgrade to patched version
-  - commit: https://github.com/projectdiscovery/nuclei/commit/1700cdf4e23128a358066364062246b58c1c013f (2024-03-14)
-- **Code Execution via Unsigned Template** (high) [CWE-347]: Introduces a flag to disable unsigned templates, preventing execution of unverified templates that could lead to arbitrary code execution.
-  - sink: `loading unsigned templates when -dut flag is set`
-  - fix: Adds -disable-unsigned-templates flag to skip unsigned templates and report them separately
-  - commit: https://github.com/projectdiscovery/nuclei/commit/fa56800fcca3c8c8fe8e8dec28584210ca1527de (2024-03-14)
-- **Auth Bypass** (medium) [CWE-347]: This commit introduces a flag to skip unsigned templates, closing a path where an attacker could execute untrusted templates by bypassing signature verification.
-  - sink: `parsed.Verified`
-  - fix: skip unverified templates when DisableUnsignedTemplates is set
-  - commit: https://github.com/projectdiscovery/nuclei/commit/9bd4db3f74aedd067af568df4f9198a8a090f68b (2024-03-10)
-- **Code Execution** (high) [CWE-94]: Prevents execution of unverified code templates in workflows, mitigating code execution risk.
-  - sink: `template.RequestsCode`
-  - fix: require -code flag and verified templates for code templates in workflows
-  - commit: https://github.com/projectdiscovery/nuclei/commit/e86f38299765b82ad724fdb701557e0eaff3884d (2024-03-01)
-- **SSRF** (high) [CWE-918]: Fixes SSRF by validating the actual LDAP controller host against the network policy instead of the domain name.
-  - sink: `newLdapSession`
-  - fix: host allowlist check on controller instead of domain
-  - commit: https://github.com/projectdiscovery/nuclei/commit/a167e6c57b80f5dc7be231231b62d7c594ee2770 (2024-01-20)
-- **Sandbox Escape** (high) [CWE-94]: Disables eval and tightens file read validation in the JavaScript sandbox, preventing template-driven code execution and file access outside the sandbox.
-  - sink: `eval function and file read in sandbox mode`
-  - fix: Disable eval function and rewrite file validation in sandbox mode
-  - commit: https://github.com/projectdiscovery/nuclei/commit/c35162c8ef061a96dd74b54c8f022c1a41cf5833 (2023-10-13)
-- **Path Traversal** (high) [CWE-22]: Adds network policy and blocks file:// protocol in headless mode unless -lfa is set, preventing local file disclosure.
-  - sink: `file:// protocol access in headless browser`
-  - fix: Network policy validation and file:// scheme blocking when local file access is disabled
-  - commit: https://github.com/projectdiscovery/nuclei/commit/d3928e080d819f25306bfbaecc399f03516db5aa (2023-08-25)
-- **Deserialization of Untrusted Data** (medium) [CWE-502]: Fixes JSON deserialization by adding explicit UnmarshalJSON methods and correcting tags, preventing type confusion and potential logic bypasses.
-  - sink: `UnmarshalJSON on user-controlled fields like Type and Name`
-  - fix: Adds proper UnmarshalJSON implementations and corrects json tags to prevent type confusion
-  - commit: https://github.com/projectdiscovery/nuclei/commit/85090b75317977bb9301444c9dfd8e4c2dba667a (2023-02-07)
-- **Sandbox Escape** (medium) [CWE-22]: The commit adds sandbox mode that restricts payload file paths to the template directory, preventing arbitrary file reads when sandbox is enabled.
-  - sink: `loadPayloadsFromFile`
-  - fix: path prefix validation against template directory and template path directory
-  - commit: https://github.com/projectdiscovery/nuclei/commit/e7fb40a41308f6c41267e858d1781eb67f7e0572 (2022-11-16)
-- **Insecure File Permissions** (low) [CWE-732]: Fixes overly permissive file permissions on debug output files by changing from os.ModePerm (0777) to 0644.
-  - sink: `os.OpenFile`
-  - fix: restrict file mode to 0644
-  - commit: https://github.com/projectdiscovery/nuclei/commit/af4854f90de728bb08689d5cae3dec4acfdb6b9a (2022-06-20)
-- **Security Misconfiguration** (medium) [CWE-16]: Prevents setting HTTP_PROXY environment variable unless explicitly enabled, avoiding potential proxy hijacking.
-  - sink: `os.Setenv(types.HTTP_PROXY_ENV, proxyURL.String())`
-  - fix: Added check for options.ProxyInternal before setting HTTP_PROXY environment variable
-  - commit: https://github.com/projectdiscovery/nuclei/commit/3f5c055aca1499642caf4f40e8cf55ad32bd4da9 (2022-03-09)
-- **Sandbox Escape** (high) [CWE-250]: The commit restricts the use of the --no-sandbox flag to Linux root users, preventing sandbox escape vulnerabilities in other environments.
-  - sink: `NoSandbox(true)`
-  - fix: Conditionally disable sandbox only on Linux as root
-  - commit: https://github.com/projectdiscovery/nuclei/commit/df55f7a2eb591a7949b5c66a9c54db54aa28f450 (2021-12-05)
-- **Path Traversal** (high) [CWE-22]: This commit fixes a zip-slip path traversal vulnerability by validating the extracted file path against the configured template directory.
-  - sink: `os.OpenFile`
-  - fix: path canonicalization and prefix check
-  - commit: https://github.com/projectdiscovery/nuclei/commit/f9c214a66f37a00fc29fe64232b690e6285ec59c (2021-11-25)
-- **SSRF** (high) [CWE-918]: Adds validation and connectivity checks for SOCKS5 proxies before launching a scan, preventing SSRF through proxy configuration.
-  - sink: `Proxy connection validation and loading`
-  - fix: Proxy server validation and connectivity check before scan launch
-  - commit: https://github.com/projectdiscovery/nuclei/commit/0e46d3e0410f89e4af41099f5cd38fa7f54ae54f (2021-11-10)
-- **Sandbox Escape** (high) [CWE-284]: This commit adds a sandbox flag that, when enabled, restricts the tengo scripting engine to safe modules, preventing sandbox escape and arbitrary command execution.
-  - sink: `stdlib.GetModuleMap(stdlib.AllModuleNames()...)`
-  - fix: Added sandbox flag to restrict imports to safe modules when sandbox is enabled
-  - commit: https://github.com/projectdiscovery/nuclei/commit/cf2a98b03d76327c20fb0183b7f1c849eb692c91 (2020-12-13)
-- **Sandbox Escape** (high) [CWE-284]: This commit restricts the tengo scripting engine to a safe module list, preventing sandbox escape and arbitrary command execution via workflow templates.
-  - sink: `stdlib.GetModuleMap(stdlib.AllModuleNames()...)`
-  - fix: Restricted tengo script imports to a safe module list
-  - commit: https://github.com/projectdiscovery/nuclei/commit/96049a32fa70aef0483d4180d386a57b3c2d973a (2020-12-13)
+**By component** (directories with the most fixes):
+
+- `pkg/protocols/common/protocolstate` — 2 fixes: SSRF (2)
+
+- [1c0f77d28e][c-1c0f77d28e] **fix(js/net): enforce network policy on proxied target in JS net dial** — SSRF, high [CWE-918] · inferred · 2026-10-18
+  `pkg/js/libs/net/net.go` · sink `dialHTTPProxy` · fix: input validation
+- [d43c1f547f][c-d43c1f547f] **fix(headless): canonicalize IPv4-mapped IPv6 literals and re-validate dial targets** — SSRF, high [CWE-918] · inferred · 2026-10-13
+  `pkg/protocols/common/protocolstate/headless.go` · sink `net.DialTimeout` · fix: canonicalization before validation
+- [d9dabbb4a1][c-d9dabbb4a1] **fix(headless): enforce network policy on ws and wss schemes (#7867)** — SSRF, high [CWE-918] · inferred · 2026-10-12
+  `pkg/protocols/common/protocolstate/headless.go` · sink `isValidHost` · fix: canonicalization before validation
+- [a20aa26024][c-a20aa26024] **Fix pprof/auth exposure on http-api-endpoint and harden file permissions** — Information Disclosure, high [CWE-200] · inferred · 2026-08-03
+  `internal/httpapi/apiendpoint.go` · sink `http.ListenAndServe` · fix: authentication verification
+- [fac0b7196d][c-fac0b7196d] **feat(templates): require signatures for javascript templates (#7514)** — Signature Verification Bypass, high [CWE-347] · verified (CVE-2026-76819, GHSA-vxg7-f2jj-jmqm) · 2026-07-03
+  `pkg/catalog/loader/loader.go` · sink `t.HasJavascriptRequest` · fix: cryptographic verification
+- [ccbfb12bd0][c-ccbfb12bd0] **refactor(protocols): template rendering boundary (#7499)** — Code Execution, high [CWE-94] · verified (CVE-2026-76805, GHSA-jpvm-9frm-hjcq) · 2026-06-30
+  `pkg/fuzz/execute.go` · sink `expressions.Evaluate` · fix: sandbox isolation
+- [3a8b65cb45][c-3a8b65cb45] **fix(js): reject unsandboxed ccache writes in krbforge (#7482)** — Sandbox Escape, high [CWE-693] · inferred · 2026-06-24
+  `pkg/js/libs/krbforge/krbforge.go` · sink `OutputFile` · fix: sandbox isolation
+- [68daef44a7][c-68daef44a7] **fix(js): escape dbname in lib/pq URL (#7479)** — SQL Injection, high [CWE-89] · inferred · 2026-06-23
+  `pkg/js/libs/postgres/postgres.go` · sink `fmt.Sprintf` (library call — the operation around it is the sink) · fix: context-aware output escaping
+- [634cecbb28][c-634cecbb28] **gate mysql allowAllFiles behind -lfa** — Arbitrary File Access, high [CWE-22] · verified (CVE-2026-76803, GHSA-xhmx-w2j4-rw3q) · 2026-06-19
+  `pkg/js/libs/mysql/mysql.go` · sink `sql.Open` · fix: sandbox isolation
+- [20cbf0bd73][c-20cbf0bd73] **refactor: use path-aware filesystem containment checks (#7420)** — Path Traversal, high [CWE-22] · inferred · 2026-06-04
+  `pkg/installer/template.go` · sink `filepath.Join` · fix: canonicalization before validation
+- [6f2ade6a9b][c-6f2ade6a9b] **fix(js): respect \`allow-local-file-access\` in \`require\` (#7332)** — Sandbox Escape, high [CWE-693] · inferred · 2026-04-10
+  `pkg/js/compiler/pool.go` · sink `require.DefaultSourceLoader` · fix: sandbox isolation
+- [d221732016][c-d221732016] **fix(expressions): only eval template-authored expressions (#7221)** — Code Execution, high [CWE-94] · verified (CVE-2026-76805, GHSA-jpvm-9frm-hjcq) · 2026-03-16
+  `pkg/protocols/common/expressions/expressions.go` · sink `FindExpressions` · fix: allowlist enforcement
+- [0da993afe6][c-0da993afe6] **Merge commit from fork** — Signature Verification Bypass, medium [CWE-347] · verified (CVE-2024-43405, GHSA-7h5p-mmpp-hgmm) · 2024-08-19
+  `pkg/templates/signer/tmpl_signer.go` · sink `os.ReadFile` · fix: cryptographic verification
+- [e5154d362a][c-e5154d362a] **fixing payload load (#3927)** — Arbitrary File Access, high [CWE-22] · verified (CVE-2023-37896, GHSA-2xx4-jj5v-6mff) · 2023-07-14
+  `v2/pkg/protocols/common/generators/load.go` · sink `loadPayloads` · fix: sandbox isolation
+
+## Historical Vulnerability Classes
+
+| Class | Fixes | Verified | Most recent fix |
+|---|---|---|---|
+| SSRF | 3 | 0 | [2026-10-18][c-1c0f77d28e] 1c0f77d28e |
+| Arbitrary File Access | 1 (1 component gone, not counted) | 2 | [2026-06-19][c-634cecbb28] 634cecbb28 |
+| Code Execution | 2 | 2 | [2026-06-30][c-ccbfb12bd0] ccbfb12bd0 |
+| Sandbox Escape | 2 | 0 | [2026-06-24][c-3a8b65cb45] 3a8b65cb45 |
+| Signature Verification Bypass | 2 | 2 | [2026-07-03][c-fac0b7196d] fac0b7196d |
+| Information Disclosure | 1 | 0 | [2026-08-03][c-a20aa26024] a20aa26024 |
+| Path Traversal | 1 | 0 | [2026-06-04][c-20cbf0bd73] 20cbf0bd73 |
+| SQL Injection | 1 | 0 | [2026-06-23][c-68daef44a7] 68daef44a7 |
+
+## Known Advisories
+
+### Linked to a fix commit in this history (5)
+
+- **CVE-2026-76805** / GHSA-jpvm-9frm-hjcq (medium 5.3) CWE-200 — fixed in [ccbfb12bd0][c-ccbfb12bd0], [d221732016][c-d221732016]
+  - Environment Variable Disclosure via Response-Derived Data in DAST/Fuzz Mode. A vulnerability in Nuclei's DAST/fuzz expression evaluation path allows a malicious target server to trigger disclosure of ...
+- **CVE-2026-76803** / GHSA-xhmx-w2j4-rw3q (medium 5.3) CWE-284 — fixed in [634cecbb28][c-634cecbb28]
+  - Local File Read via MySQL Client Sandbox Bypass. A vulnerability in Nuclei's JavaScript MySQL client library allows arbitrary local file reads that bypass the `-allow-local-file-access` (`-lfa`) sandb...
+- **CVE-2026-76819** / GHSA-vxg7-f2jj-jmqm (high 8.6) CWE-94,CWE-787 — fixed in [fac0b7196d][c-fac0b7196d]
+  - Arbitrary Code Execution via Goja JavaScript Engine Vulnerability. A vulnerability in the Goja JavaScript engine used by Nuclei's `javascript:` protocol allows arbitrary native code execution on the s...
+- **CVE-2024-43405** / GHSA-7h5p-mmpp-hgmm (high 7.4) CWE-78 — fixed in [0da993afe6][c-0da993afe6]
+  - Nuclei Template Signature Verification Bypass. ## Summary A vulnerability has been identified in Nuclei's template signature verification system that could allow an attacker to bypass the signature ch...
+- **CVE-2023-37896** / GHSA-2xx4-jj5v-6mff (medium 5.9) CWE-22 — fixed in [e5154d362a][c-e5154d362a]
+  - Nuclei < 2.9.9 - Sandbox restriction bypass. ## Overview Affected: < 2.9.9 (patched 2.9.9).
+
+### Without a linked fix commit (6)
+
+Published elsewhere, fixed before the history window, or not yet linked; full descriptions are in the JSON context.
+
+| Advisory | Severity | Published | CWE | Summary |
+|---|---|---|---|---|
+| CVE-2026-76804 / GHSA-qgw5-7j4f-fg97 | medium 5.5 | 2026-07-16 | CWE-284 | Local File Read via Workflow File-Protocol Gate Bypass. A vulnerability in Nuclei's workfl... |
+| CVE-2026-76802 / GHSA-jpf4-98qj-qr67 | medium 4.7 | 2026-07-16 | CWE-78 | Arbitrary Command Execution via DAST Code Signature Bypass. A vulnerability in Nuclei's DA... |
+| CVE-2026-41645 / GHSA-jm34-66cf-qpvr | medium 5.3 | 2026-04-18 | CWE-94 | Environment variable disclosure via Response-Derived DSL Expressions. A vulnerability in N... |
+| CVE-2026-41646 / GHSA-29rg-wmcw-hpf4 | medium 5.5 | 2026-04-18 | CWE-284 | Local File Read via require() Module Loader Bypass. A vulnerability in Nuclei's JavaScript... |
+| CVE-2024-40641 / GHSA-c3q9-c27p-cw9h | high 7.4 | 2024-07-17 | CWE-78 | Unsigned code template execution through workflows in projectdiscovery/nuclei. ### Summary... |
+| CVE-2024-27920 / GHSA-w5wx-6g2r-r78q | high 7.4 | 2024-03-15 | CWE-78 | Unsigned code template execution through workflows in projectdiscovery/nuclei. ### Overvie... |
+
+[c-df57ae6d29]: https://github.com/projectdiscovery/nuclei/commit/df57ae6d2955864fd92a1bd2565fe9e4e9a7592b
+[c-d43c1f547f]: https://github.com/projectdiscovery/nuclei/commit/d43c1f547f98ac56654f4335ad20af0cca4c229c
+[c-d9dabbb4a1]: https://github.com/projectdiscovery/nuclei/commit/d9dabbb4a19e897a6bc12334a030a43035b27b73
+[c-e5154d362a]: https://github.com/projectdiscovery/nuclei/commit/e5154d362af900e83b30fb79028bfab1a7f0c648
+[c-634cecbb28]: https://github.com/projectdiscovery/nuclei/commit/634cecbb289db494247c04a9c4ec6a113aef3cc0
+[c-ccbfb12bd0]: https://github.com/projectdiscovery/nuclei/commit/ccbfb12bd01447ba25f6e755dfb2bee677736da6
+[c-d221732016]: https://github.com/projectdiscovery/nuclei/commit/d2217320162d5782ca7cb95bef9dda17063818f3
+[c-a20aa26024]: https://github.com/projectdiscovery/nuclei/commit/a20aa260245e5db889c5a421d4a65aeb2a64bf46
+[c-1c0f77d28e]: https://github.com/projectdiscovery/nuclei/commit/1c0f77d28ed639a34c2036fa23c7503ffac0bf27
+[c-fac0b7196d]: https://github.com/projectdiscovery/nuclei/commit/fac0b7196d2fbca56c8af8ff8e1899c111041a37
+[c-3a8b65cb45]: https://github.com/projectdiscovery/nuclei/commit/3a8b65cb45f121fcdf418f60e5177ed9f250c122
+[c-68daef44a7]: https://github.com/projectdiscovery/nuclei/commit/68daef44a7eafd41c39690924620e418762909ef
+[c-20cbf0bd73]: https://github.com/projectdiscovery/nuclei/commit/20cbf0bd73893c1270582c9778b6b26a5cb11206
+[c-6f2ade6a9b]: https://github.com/projectdiscovery/nuclei/commit/6f2ade6a9b427c284c15a43445f9c7f055e60e5d
+[c-0da993afe6]: https://github.com/projectdiscovery/nuclei/commit/0da993afe6d41b4b1b814e8fad23a2acba13c60a
 
 ---
-_Fixed-issue fingerprints and CVE matches are derived from real data; LLM-written prose is advisory and should be verified against the cited commits._
+_Fixed-issue fingerprints and advisory matches are derived from git history and published advisories; the risk tables and pattern descriptions above are derived from them and should be verified against the cited commits._

SYNTAX-REFERENCE.md

@@ -530,6 +530,16 @@ Description of the template.
 
 You can go in-depth here on what the template actually does.
 
+Description, impact and remediation may hold {{placeholders}} filled for
+each finding: named extractors, payloads, variables and DSL helpers
+(network helpers excluded), and the result fields host, port, ip, url,
+scheme, path, matched-at, matcher-name and extractor-name, which win over
+values of the same name. Use join(", ", name) for an extractor with
+several values, and brackets for a hyphenated name inside a helper:
+to_upper([matcher-name]). Unknown placeholders stay as written, and so
+does a field whose expression fails. Global matcher templates match on
+other templates' responses, so only the result fields fill their text.
+
 
 
 Examples:
@@ -543,6 +553,10 @@ description: Bower is a package manager which stores package information in the
 description: Subversion ALM for the enterprise before 8.8.2 allows reflected XSS at multiple locations
 ```
 
+```yaml
+description: Redis {{version}} on {{host}}:{{port}} accepts the default password
+```
+
 
 </div>
 
@@ -559,6 +573,8 @@ Impact of the template.
 
 You can go in-depth here on impact of the template.
 
+Supports the same placeholders as description.
+
 
 
 Examples:
@@ -665,6 +681,8 @@ Remediation steps for the template.
 
 You can go in-depth here on how to mitigate the problem found by this template.
 
+Supports the same placeholders as description.
+
 
 
 Examples:
@@ -3757,6 +3775,8 @@ Part Definitions:
 - <code>req</code> - Headless request made from the client
 - <code>duration</code> - Latest measured operation duration in seconds
 - <code>resp,body,data</code> - Headless response received from client (default)
+- <code>header,all_headers</code> - Headless response headers
+- <code>status_code</code> - Status code received from the server
 
 <hr />
 

_typos.toml

@@ -20,6 +20,8 @@ extend-exclude = [
     "vendor/",
     # Go checksum file
     "go.sum",
+    # Generated history. Commit SHAs are not words.
+    "SECURITY_CONTEXT.md",
 ]
 
 [default.extend-identifiers]

cmd/nuclei/issue-tracker-config.yaml

@@ -57,7 +57,8 @@
 #    severity: low
 #  # duplicate-issue-check (optional) flag to enable duplicate tracking issue check
 #  duplicate-issue-check: false
-#  # duplicate-issue-page-size (optional) controls how many issues to fetch per page when searching for duplicates
+#  # duplicate-issue-page-size (optional) sets the number of issues to request per page when searching for duplicates
+#  # (GitLab returns at most 100 issues per page)
 #  duplicate-issue-page-size: 100
 #  # duplicate-issue-max-pages (optional) limits how many pages to fetch when searching for duplicates (0 = no limit)
 #  duplicate-issue-max-pages: 0

cmd/nuclei/main.go

@@ -271,10 +271,10 @@ on extensive configurability, massive extensibility and ease of use.`)
 		flagSet.StringSliceVarP(&options.Targets, "target", "u", nil, "target URLs/hosts to scan", goflags.CommaSeparatedStringSliceOptions),
 		flagSet.StringVarP(&options.TargetsFilePath, "list", "l", "", "path to file containing a list of target URLs/hosts to scan (one per line)"),
 		flagSet.StringVarP(&options.InlineTargetsList, "targets-inline", "", "", "inline multiline target list (for use in template profiles)"),
-		flagSet.StringSliceVarP(&options.ExcludeTargets, "exclude-hosts", "eh", nil, "hosts to exclude to scan from the input list (ip, cidr, hostname)", goflags.FileCommaSeparatedStringSliceOptions),
+		flagSet.StringSliceVarP(&options.ExcludeTargets, "exclude-hosts", "eh", nil, "hosts to exclude to scan from the input list (ip, cidr, hostname)", goflags.FileNormalizedOriginalStringSliceOptions),
 		flagSet.StringVar(&options.Resume, "resume", "", "resume scan from and save to specified file (clustering will be disabled)"),
 		flagSet.BoolVarP(&options.ScanAllIPs, "scan-all-ips", "sa", false, "scan all the IP's associated with dns record"),
-		flagSet.StringSliceVarP(&options.IPVersion, "ip-version", "iv", nil, "IP version to scan of hostname (4,6) - (default 4)", goflags.CommaSeparatedStringSliceOptions),
+		flagSet.StringSliceVarP(&options.IPVersion, "ip-version", "iv", nil, "IP version to scan of hostname (4,6); default uses dialer fallback, -sa scans all addresses", goflags.CommaSeparatedStringSliceOptions),
 		flagSet.BoolVarP(&options.DisableURLNormalization, "disable-url-normalization", "dun", false, "keep targets exactly as supplied instead of collapsing urls that address the same resource"),
 		flagSet.IntVarP(&options.MaxURLsPerPattern, "max-urls-per-pattern", "mup", 0, "scan at most this many urls sharing one structural shape, e.g. /user/1 and /user/2 (0 to scan all)"),
 	)
@@ -309,7 +309,7 @@ on extensive configurability, massive extensibility and ease of use.`)
 		flagSet.BoolVarP(&options.EnableGlobalMatchersTemplates, "enable-global-matchers", "egm", false, "enable loading global matchers templates"),
 		flagSet.BoolVar(&options.EnableFileTemplates, "file", false, "enable loading file templates"),
 		flagSet.BoolVar(&options.EnableLLM, "llm", false, "enable llm matchers and extractors (semantic matching)"),
-		flagSet.StringVar(&options.LLMProvider, "llm-provider", "openai", "llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter)"),
+		flagSet.StringVar(&options.LLMProvider, "llm-provider", "openai", "llm provider for semantic matching (openai, ollama, llamacpp, vllm, lmstudio, groq, openrouter, together, orcarouter, jev)"),
 		flagSet.StringVar(&options.LLMBaseURL, "llm-base-url", "", "openai-compatible endpoint for semantic matching (overrides -llm-provider, e.g. a local model)"),
 		flagSet.StringVar(&options.LLMModel, "llm-model", "", "model used for semantic matching"),
 		flagSet.IntVar(&options.LLMTimeout, "llm-timeout", 30, "time in seconds to wait for a single llm call"),
@@ -355,6 +355,7 @@ on extensive configurability, massive extensibility and ease of use.`)
 		flagSet.StringVarP(&options.JSONLExport, "jsonl-export", "jle", "", "file to export results in JSONL(ine) format"),
 		flagSet.StringVarP(&options.PDFExport, "pdf-export", "pe", "", "file to export results in PDF format"),
 		flagSet.StringSliceVarP(&options.Redact, "redact", "rd", nil, "redact given list of keys from query parameter, request header and body", goflags.CommaSeparatedStringSliceOptions),
+		flagSet.BoolVarP(&options.NoRedact, "no-redact", "nrd", false, "disable the built-in redaction of credential headers (Authorization, Cookie, Set-Cookie, ...) in output; keys given with -redact are still redacted"),
 	)
 
 	flagSet.CreateGroup("configs", "Configurations",
@@ -470,6 +471,7 @@ on extensive configurability, massive extensibility and ease of use.`)
 		flagSet.BoolVar(&options.PreflightPortScan, "preflight-portscan", false, "run preflight resolve + TCP portscan and filter targets before scanning (disabled by default)"),
 		flagSet.BoolVar(&options.DisableStdin, "no-stdin", false, "disable stdin processing"),
 		flagSet.BoolVar(&options.EnableHTTPCache, "http-cache", false, "enable HTTP cache (RFC 9111) for HTTP requests"),
+		flagSet.BoolVarP(&options.AutoCalibrate, "auto-calibrate", "ac", false, "drop HTTP matches on a host's catch-all page (answers missing paths with 2xx/3xx)"),
 	)
 
 	flagSet.CreateGroup("headless", "Headless",

go.mod

@@ -18,10 +18,11 @@ require (
 	github.com/alecthomas/chroma v0.10.0
 	github.com/alexsnet/go-vnc v0.1.0
 	github.com/alitto/pond v1.9.2
+	github.com/andybalholm/brotli v1.2.0
 	github.com/andygrunwald/go-jira v1.16.1
-	github.com/antchfx/htmlquery v1.3.5
+	github.com/antchfx/htmlquery v1.3.7
 	github.com/antchfx/xmlquery v1.4.4
-	github.com/antchfx/xpath v1.3.6
+	github.com/antchfx/xpath v1.3.9
 	github.com/aws/aws-sdk-go-v2 v1.41.5
 	github.com/aws/aws-sdk-go-v2/config v1.31.16
 	github.com/aws/aws-sdk-go-v2/credentials v1.18.20
@@ -33,10 +34,11 @@ require (
 	github.com/clbanning/mxj/v2 v2.7.0
 	github.com/ditashi/jsbeautifier-go v0.0.0-20141206144643-2520a8026a9c
 	github.com/docker/go-units v0.5.0
+	github.com/dsnet/compress v0.0.2-0.20230904184137-39efe44ab707
 	github.com/elazarl/goproxy v1.7.2
 	github.com/fatih/structs v1.1.0
 	github.com/fullstorydev/grpcurl v1.9.3
-	github.com/getkin/kin-openapi v0.144.0
+	github.com/getkin/kin-openapi v0.149.0
 	github.com/getsops/sops/v3 v3.9.2
 	github.com/go-echarts/go-echarts/v2 v2.6.0
 	github.com/go-git/go-git/v5 v5.19.2
@@ -61,6 +63,7 @@ require (
 	github.com/jhump/protoreflect v1.17.0
 	github.com/julienschmidt/httprouter v1.3.0
 	github.com/kitabisa/go-ci v1.0.3
+	github.com/klauspost/compress v1.18.7
 	github.com/leslie-qiwa/flat v0.0.0-20230424180412-f9d1cf014baa
 	github.com/lib/pq v1.12.3
 	github.com/logrusorgru/aurora/v4 v4.0.0
@@ -75,7 +78,7 @@ require (
 	github.com/pkg/errors v0.9.1
 	github.com/praetorian-inc/fingerprintx v1.1.15
 	github.com/projectdiscovery/clistats v0.1.7
-	github.com/projectdiscovery/dsl v0.8.24
+	github.com/projectdiscovery/dsl v0.9.0
 	github.com/projectdiscovery/fastdialer v0.6.0
 	github.com/projectdiscovery/fasttemplate v0.0.2
 	github.com/projectdiscovery/gcache v0.0.0-20241015120333-12546c6e3f4c
@@ -85,7 +88,7 @@ require (
 	github.com/projectdiscovery/goja_nodejs v0.0.0-20260618132410-8519f75f703d
 	github.com/projectdiscovery/gologger v1.1.73
 	github.com/projectdiscovery/gostruct v0.0.2
-	github.com/projectdiscovery/govaluate v0.0.0-20260615100919-5ee2581bbf7e
+	github.com/projectdiscovery/govaluate v0.0.0-20261102102118-9294133b15a2
 	github.com/projectdiscovery/gozero v0.1.1-0.20260530071156-fa1dad563d76
 	github.com/projectdiscovery/hmap v0.0.102
 	github.com/projectdiscovery/httpx v1.12.0
@@ -102,8 +105,8 @@ require (
 	github.com/projectdiscovery/tlsx v1.4.3-0.20261020052910-acb80c2c7c75
 	github.com/projectdiscovery/uncover v1.2.1
 	github.com/projectdiscovery/useragent v0.0.109
-	github.com/projectdiscovery/utils v0.11.12-0.20261025172519-64b3c221cfe5
-	github.com/projectdiscovery/wappalyzergo v0.3.5
+	github.com/projectdiscovery/utils v0.11.13-0.20261103094816-5c0b20e35350
+	github.com/projectdiscovery/wappalyzergo v0.3.6
 	github.com/projectdiscovery/yamldoc-go v1.0.7
 	github.com/redis/go-redis/v9 v9.19.0
 	github.com/remeh/sizedwaitgroup v1.0.0
@@ -113,6 +116,7 @@ require (
 	github.com/segmentio/ksuid v1.0.4
 	github.com/shurcooL/graphql v0.0.0-20230722043721-ed46e5a46466
 	github.com/sijms/go-ora/v2 v2.9.0
+	github.com/spaolacci/murmur3 v1.1.0
 	github.com/spf13/cast v1.10.0
 	github.com/stretchr/testify v1.12.1
 	github.com/syndtr/goleveldb v1.0.0
@@ -175,7 +179,6 @@ require (
 	github.com/alecthomas/kingpin/v2 v2.4.0 // indirect
 	github.com/alecthomas/template v0.0.0-20190718012654-fb15b899a751 // indirect
 	github.com/alecthomas/units v0.0.0-20211218093645-b94a6e3cc137 // indirect
-	github.com/andybalholm/brotli v1.2.0 // indirect
 	github.com/andybalholm/cascadia v1.3.4 // indirect
 	github.com/asaskevich/govalidator v0.0.0-20230301143203-a9d515a09cc2 // indirect
 	github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.8 // indirect
@@ -239,7 +242,6 @@ require (
 	github.com/docker/cli v29.2.0+incompatible // indirect
 	github.com/docker/go-connections v0.7.0 // indirect
 	github.com/dop251/goja v0.0.0-20260311135729-065cd970411c // indirect
-	github.com/dsnet/compress v0.0.2-0.20230904184137-39efe44ab707 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/ebitengine/purego v0.10.0 // indirect
 	github.com/emersion/go-sasl v0.0.0-20241020182733-b788ff22d5a6 // indirect
@@ -323,7 +325,6 @@ require (
 	github.com/kaiakz/ubuffer v0.0.0-20200803053910-dd1083087166 // indirect
 	github.com/kataras/jwt v0.1.10 // indirect
 	github.com/kevinburke/ssh_config v1.2.0 // indirect
-	github.com/klauspost/compress v1.18.7 // indirect
 	github.com/klauspost/cpuid/v2 v2.3.0 // indirect
 	github.com/klauspost/pgzip v1.2.6 // indirect
 	github.com/kylelemons/godebug v1.1.0 // indirect
@@ -385,14 +386,13 @@ require (
 	github.com/russross/blackfriday/v2 v2.1.0 // indirect
 	github.com/ryanuber/go-glob v1.0.0 // indirect
 	github.com/saintfish/chardet v0.0.0-20230101081208-5e3ef4b5456d // indirect
-	github.com/santhosh-tekuri/jsonschema/v6 v6.0.2 // indirect
+	github.com/santhosh-tekuri/jsonschema/v6 v6.0.3 // indirect
 	github.com/sergi/go-diff v1.4.0 // indirect
 	github.com/shirou/gopsutil v3.21.11+incompatible // indirect
 	github.com/shirou/gopsutil/v4 v4.26.3 // indirect
 	github.com/sirupsen/logrus v1.10.2 // indirect
 	github.com/skeema/knownhosts v1.3.1 // indirect
 	github.com/sorairolake/lzip-go v0.3.8 // indirect
-	github.com/spaolacci/murmur3 v1.1.0 // indirect
 	github.com/spf13/afero v1.15.0 // indirect
 	github.com/spf13/pflag v1.0.10 // indirect
 	github.com/spiffe/go-spiffe/v2 v2.7.0 // indirect

go.sum

@@ -180,14 +180,13 @@ github.com/andygrunwald/go-jira v1.16.1 h1:WoQEar5XoDRAibOgKzTFELlPNlKAtnfWr296R
 github.com/andygrunwald/go-jira v1.16.1/go.mod h1:UQH4IBVxIYWbgagc0LF/k9FRs9xjIiQ8hIcC6HfLwFU=
 github.com/anmitsu/go-shlex v0.0.0-20200514113438-38f4b401e2be h1:9AeTilPcZAjCFIImctFaOjnTIavg87rW78vTPkQqLI8=
 github.com/anmitsu/go-shlex v0.0.0-20200514113438-38f4b401e2be/go.mod h1:ySMOLuWl6zY27l47sB3qLNK6tF2fkHG55UZxx8oIVo4=
-github.com/antchfx/htmlquery v1.3.5 h1:aYthDDClnG2a2xePf6tys/UyyM/kRcsFRm+ifhFKoU0=
-github.com/antchfx/htmlquery v1.3.5/go.mod h1:5oyIPIa3ovYGtLqMPNjBF2Uf25NPCKsMjCnQ8lvjaoA=
+github.com/antchfx/htmlquery v1.3.7 h1:KCUklE1EO/xm6iS3CWx6iatFlcnENBk4I4UYa2pBmLU=
+github.com/antchfx/htmlquery v1.3.7/go.mod h1:/uBTMnFEl6LaBlIgw3KZ+c061tAQdlfbCXObZoxwQf4=
 github.com/antchfx/xmlquery v1.4.4 h1:mxMEkdYP3pjKSftxss4nUHfjBhnMk4imGoR96FRY2dg=
 github.com/antchfx/xmlquery v1.4.4/go.mod h1:AEPEEPYE9GnA2mj5Ur2L5Q5/2PycJ0N9Fusrx9b12fc=
 github.com/antchfx/xpath v1.3.3/go.mod h1:i54GszH55fYfBmoZXapTHN8T8tkcHfRgLyVwwqzXNcs=
-github.com/antchfx/xpath v1.3.5/go.mod h1:i54GszH55fYfBmoZXapTHN8T8tkcHfRgLyVwwqzXNcs=
-github.com/antchfx/xpath v1.3.6 h1:s0y+ElRRtTQdfHP609qFu0+c6bglDv20pqOViQjjdPI=
-github.com/antchfx/xpath v1.3.6/go.mod h1:i54GszH55fYfBmoZXapTHN8T8tkcHfRgLyVwwqzXNcs=
+github.com/antchfx/xpath v1.3.9 h1:muXJSod6jRzhsiUjvduNSxsRSptFZ2GOJ1oMD1QkPSA=
+github.com/antchfx/xpath v1.3.9/go.mod h1:i54GszH55fYfBmoZXapTHN8T8tkcHfRgLyVwwqzXNcs=
 github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5 h1:0CwZNZbxp69SHPdPJAN/hZIm0C4OItdklCFmMRWYpio=
 github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5/go.mod h1:wHh0iHkYZB8zMSxRWpUBQtwG5a7fFgvEO+odwuTv2gs=
 github.com/asaskevich/govalidator v0.0.0-20230301143203-a9d515a09cc2 h1:DklsrG3dyBCFEj5IhUbnKptjxatkF07cF2ak3yi77so=
@@ -427,8 +426,8 @@ github.com/gaissmai/bart v0.30.0 h1:K1oTSQ6Nf76bhrgxeTqzZa4hE0M9x4xDfL6n6UG67Sc=
 github.com/gaissmai/bart v0.30.0/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
 github.com/geoffgarside/ber v1.2.0 h1:/loowoRcs/MWLYmGX9QtIAbA+V/FrnVLsMMPhwiRm64=
 github.com/geoffgarside/ber v1.2.0/go.mod h1:jVPKeCbj6MvQZhwLYsGwaGI52oUorHoHKNecGT85ZCc=
-github.com/getkin/kin-openapi v0.144.0 h1:hIRcTH+KjLfkLpYU6bSSfdFpi0fZi1fp+hSPi4aQu9Y=
-github.com/getkin/kin-openapi v0.144.0/go.mod h1:3BH9M9XDe/y9M5DSvEocVYAYq1w0qrhJHjC/vZi0AaY=
+github.com/getkin/kin-openapi v0.149.0 h1:ZbhmVJ4yq5RZDUsyP8lcBcGMsjsaTqXEFt6isdtMDfA=
+github.com/getkin/kin-openapi v0.149.0/go.mod h1:1+BHDzstro+P5CKtPy1X4PfofnFgmRe6uvMy9+r9fKY=
 github.com/getsops/gopgagent v0.0.0-20240527072608-0c14999532fe h1:QKe/kmAYbndxwu91TcjHERsnMh5SgOB1x/qicvOdUJ8=
 github.com/getsops/gopgagent v0.0.0-20240527072608-0c14999532fe/go.mod h1:awFzISqLJoZLm+i9QQ4SgMNHDqljH6jWV0B36V5MrUM=
 github.com/getsops/sops/v3 v3.9.2 h1:Cuuahc/UGu7W3+NXd+mrl+2oXGxxj/jIPsw6Ah2CbxA=
@@ -935,8 +934,8 @@ github.com/projectdiscovery/cdncheck v1.3.2 h1:I453ZWNBeG0WWIy3wi9412l4PTzorLgeb
 github.com/projectdiscovery/cdncheck v1.3.2/go.mod h1:5IkUMMlcGaeg1DWW5oJVd3r9X3EpJ88j5HxBeF4Me9M=
 github.com/projectdiscovery/clistats v0.1.7 h1:NXf4lUzx1QcbJd6QycNwwYHTv2FsW9MchQ2LsgSL6C0=
 github.com/projectdiscovery/clistats v0.1.7/go.mod h1:RxRGxhMzkT344VTTvZWIrojurUXdQQRN1yfwgg/88I8=
-github.com/projectdiscovery/dsl v0.8.24 h1:rN2w6LSIjA16qKGRjyj70oxdwXGaspe6sAjEkP7F2fk=
-github.com/projectdiscovery/dsl v0.8.24/go.mod h1:V+EludvS5PS19UNMlM3CuBwZ+sFnMaQhCFGxysDUanQ=
+github.com/projectdiscovery/dsl v0.9.0 h1:i1KyCbKJyKfHkgdbfh0JkWdd14APUjpEYf+GaYANlCg=
+github.com/projectdiscovery/dsl v0.9.0/go.mod h1:ntNFkkiYCTEosGFTbApA61RMrvJTCjnakPxnWj56MyY=
 github.com/projectdiscovery/fastdialer v0.6.0 h1:4ZeFQ6LyRH4D8e3qO/yzqIP4Cf9Fr0DduaVpJD6ENnE=
 github.com/projectdiscovery/fastdialer v0.6.0/go.mod h1:FpHwa9S5PUgfPbh+irOm+cdU6AO631XidxkhPwPv7yY=
 github.com/projectdiscovery/fasttemplate v0.0.2 h1:h2cISk5xDhlJEinlBQS6RRx0vOlOirB2y3Yu4PJzpiA=
@@ -957,8 +956,8 @@ github.com/projectdiscovery/gologger v1.1.73 h1:WX2nzq29cL7HdvOHqIttacoid/5lMwap
 github.com/projectdiscovery/gologger v1.1.73/go.mod h1:mJwODZcFDg70ihINpOvZevmBtgvpP8H9/l8Y+OPhZPY=
 github.com/projectdiscovery/gostruct v0.0.2 h1:s8gP8ApugGM4go1pA+sVlPDXaWqNP5BBDDSv7VEdG1M=
 github.com/projectdiscovery/gostruct v0.0.2/go.mod h1:H86peL4HKwMXcQQtEa6lmC8FuD9XFt6gkNR0B/Mu5PE=
-github.com/projectdiscovery/govaluate v0.0.0-20260615100919-5ee2581bbf7e h1:vxzgQlz2Cy/YvizYDQx9OhucBcmBotfDhbQ4yCY2vfA=
-github.com/projectdiscovery/govaluate v0.0.0-20260615100919-5ee2581bbf7e/go.mod h1:xH7bPwHxUlz1yx9UlVeTF+UVCUaKhTnZgaxHb5z362E=
+github.com/projectdiscovery/govaluate v0.0.0-20261102102118-9294133b15a2 h1:aSLbDvrJQPlp4HvojU+bTU5XV8xdAVCjoc5gX1Wfbw4=
+github.com/projectdiscovery/govaluate v0.0.0-20261102102118-9294133b15a2/go.mod h1:Z3VjpUJV1pnXDwvAddA8aZ4e839SoeftB7CSdZntAVI=
 github.com/projectdiscovery/gozero v0.1.1-0.20260530071156-fa1dad563d76 h1:AN70bbi6BBs7KpIM9w0LxygUN7uzT/oH+owDIQ+Fz/k=
 github.com/projectdiscovery/gozero v0.1.1-0.20260530071156-fa1dad563d76/go.mod h1:cWHYnRXoYWHtTpOYyAp5laGYX8GH8ITUhgQaP8G/8FA=
 github.com/projectdiscovery/hmap v0.0.102 h1:ybPuf7UQvuOLiHLOUbVMvd+PSlItQU5iQpP1Erl8kaU=
@@ -999,10 +998,10 @@ github.com/projectdiscovery/uncover v1.2.1 h1:8U46T/96CLT7BPoXBgkTvWqB06lOyeTSLv
 github.com/projectdiscovery/uncover v1.2.1/go.mod h1:0p8onrWxfpXQEYs90ZDzTSpu1107gWmodX1NWqu/+z4=
 github.com/projectdiscovery/useragent v0.0.109 h1:b86BPdZvmgbJousa04jAugDRY4Y+rAvEamE0TFCPmeE=
 github.com/projectdiscovery/useragent v0.0.109/go.mod h1:sRyBqDAcD31RKYrO8fyXGowLd36fNwY1g0MNW9mp7zk=
-github.com/projectdiscovery/utils v0.11.12-0.20261025172519-64b3c221cfe5 h1:AOspf7ScbwZr8A2YnjCFd4ZJAMSFqzfEaIJIbJPuu10=
-github.com/projectdiscovery/utils v0.11.12-0.20261025172519-64b3c221cfe5/go.mod h1:TJgsyVjB02CavwZVFzUzPg6YOSJXQxLhL808JCiQKlU=
-github.com/projectdiscovery/wappalyzergo v0.3.5 h1:UZradYF1GIcruSIV5wwo/Coph9/VnM17oajLD9eXR2U=
-github.com/projectdiscovery/wappalyzergo v0.3.5/go.mod h1:4E6+fh/ckaY3wo5tlM7lNYSkhuTYRHzu2YmGIRraXP0=
+github.com/projectdiscovery/utils v0.11.13-0.20261103094816-5c0b20e35350 h1:uXqpYa6XKV9dCvoti9PbSgIWIsBKPu5YKMywB9G4q0Q=
+github.com/projectdiscovery/utils v0.11.13-0.20261103094816-5c0b20e35350/go.mod h1:TJgsyVjB02CavwZVFzUzPg6YOSJXQxLhL808JCiQKlU=
+github.com/projectdiscovery/wappalyzergo v0.3.6 h1:1FXCMeCa1xVu2zaW+eeBcbVxtlMStqGOlKcMLyh2pzI=
+github.com/projectdiscovery/wappalyzergo v0.3.6/go.mod h1:4E6+fh/ckaY3wo5tlM7lNYSkhuTYRHzu2YmGIRraXP0=
 github.com/projectdiscovery/yamldoc-go v1.0.7 h1:Odmjd1o4EkMff2F6QHi1dIYPIQPZUFy24YqmQCprmTk=
 github.com/projectdiscovery/yamldoc-go v1.0.7/go.mod h1:nqD1iPMcofpRA1DbXavPByJhWURrshT+l8WN9XjyQag=
 github.com/prometheus/client_golang v0.9.1/go.mod h1:7SWBe2y4D6OKWSNQJUaRYU/AaXPKyh/dDVn+NZz0KFw=
@@ -1055,8 +1054,8 @@ github.com/saintfish/chardet v0.0.0-20230101081208-5e3ef4b5456d h1:hrujxIzL1woJ7
 github.com/saintfish/chardet v0.0.0-20230101081208-5e3ef4b5456d/go.mod h1:uugorj2VCxiV1x+LzaIdVa9b4S4qGAcH6cbhh4qVxOU=
 github.com/sandrolain/httpcache v1.4.0 h1:Jf4Vx62X2ybvNPSpPvI1kT3xvMdDG1AsApQjOQKO9E0=
 github.com/sandrolain/httpcache v1.4.0/go.mod h1:kHBuXveitSn39SNPBhdf/ybG272X706HJ2RJqOQ+Em0=
-github.com/santhosh-tekuri/jsonschema/v6 v6.0.2 h1:KRzFb2m7YtdldCEkzs6KqmJw4nqEVZGK7IN2kJkjTuQ=
-github.com/santhosh-tekuri/jsonschema/v6 v6.0.2/go.mod h1:JXeL+ps8p7/KNMjDQk3TCwPpBy0wYklyWTfbkIzdIFU=
+github.com/santhosh-tekuri/jsonschema/v6 v6.0.3 h1:1EYB5IzjZawrrnELUi78f9fPu57HuXjmddZPjrls/28=
+github.com/santhosh-tekuri/jsonschema/v6 v6.0.3/go.mod h1:JXeL+ps8p7/KNMjDQk3TCwPpBy0wYklyWTfbkIzdIFU=
 github.com/segmentio/ksuid v1.0.4 h1:sBo2BdShXjmcugAMwjugoGUdUV0pcxY5mW4xKRn3v4c=
 github.com/segmentio/ksuid v1.0.4/go.mod h1:/XUiZBD3kVx5SmUOl55voK5yeAbBNNIed+2O73XgrPE=
 github.com/sergi/go-diff v1.4.0 h1:n/SP9D5ad1fORl+llWyN+D6qoUETXNZARKjyY2/KVCw=

internal/runner/inputs.go

@@ -39,6 +39,9 @@ func (r *Runner) initializeTemplatesHTTPInput() (*hybrid.HybridMap, error) {
 	}
 	httpxOptions.RetryMax = r.options.Retries
 	httpxOptions.Timeout = time.Duration(r.options.Timeout) * time.Second
+	if len(r.options.InternalResolversList) > 0 {
+		httpxOptions.Resolvers = r.options.InternalResolversList
+	}
 
 	dialers := protocolstate.GetDialersWithId(r.options.ExecutionId)
 	if dialers == nil {

internal/runner/llm.go

@@ -2,8 +2,12 @@ package runner
 
 import (
 	"context"
+	"errors"
+	"sync"
 	"time"
 
+	"github.com/projectdiscovery/gologger"
+
 	llmclient "github.com/projectdiscovery/nuclei/v3/pkg/operators/common/llm"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
 	utilsllm "github.com/projectdiscovery/utils/llm"
@@ -28,6 +32,32 @@ func (a *llmMatcherClient) Complete(ctx context.Context, prompt string, asJSON b
 	})
 }
 
+// llmJudgeClient is an llmMatcherClient whose provider classifies natively, so
+// llm matchers ask it through Judge instead of Complete.
+type llmJudgeClient struct {
+	*llmMatcherClient
+	warnComplete sync.Once
+}
+
+func (a *llmJudgeClient) Judge(ctx context.Context, question string, labels, inputs []string) (llmclient.Verdict, error) {
+	verdict, err := a.client.Judge(ctx, utilsllm.JudgeRequest{Question: question, Labels: labels, Inputs: inputs})
+	if err != nil {
+		return llmclient.Verdict{}, err
+	}
+	return llmclient.Verdict{Label: verdict.Label, Confidence: verdict.Confidence, Model: verdict.Model}, nil
+}
+
+// Complete is only reached by llm extractors, which need free text a
+// classifier cannot produce. Returning here spends no call budget and
+// makes no provider request. Extractors turn the error into an empty
+// result, and the warning says why.
+func (a *llmJudgeClient) Complete(context.Context, string, bool) (string, error) {
+	a.warnComplete.Do(func() {
+		gologger.Warning().Msgf("llm provider %q cannot run llm extractors, they return nothing", utilsllm.ProviderJev)
+	})
+	return "", errors.New("llm provider cannot run extractors")
+}
+
 // configureLLM builds the scan's llm client from options. It returns nil unless
 // -llm is set, so nothing reaches a model until the user opts in. The API key is
 // read from the environment by the provider layer, never from options.
@@ -52,5 +82,9 @@ func configureLLM(options *types.Options) (llmclient.Client, error) {
 		return nil, err
 	}
 
-	return &llmMatcherClient{client: client, model: options.LLMModel}, nil
+	matcherClient := &llmMatcherClient{client: client, model: options.LLMModel}
+	if client.CanJudge() {
+		return &llmJudgeClient{llmMatcherClient: matcherClient}, nil
+	}
+	return matcherClient, nil
 }

internal/runner/llm_test.go

@@ -0,0 +1,65 @@
+package runner
+
+import (
+	"context"
+	"net/http"
+	"net/http/httptest"
+	"sync/atomic"
+	"testing"
+
+	llmclient "github.com/projectdiscovery/nuclei/v3/pkg/operators/common/llm"
+	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/stretchr/testify/require"
+)
+
+func TestConfigureLLMExposesJudgeOnlyForClassifiers(t *testing.T) {
+	t.Setenv("LLM_API_KEY", "key")
+
+	client, err := configureLLM(&types.Options{EnableLLM: true, LLMProvider: "jev", LLMModel: "jev-latest"})
+	require.NoError(t, err)
+	require.Implements(t, (*llmclient.Judge)(nil), client)
+
+	client, err = configureLLM(&types.Options{EnableLLM: true, LLMProvider: "ollama", LLMModel: "llama3.2"})
+	require.NoError(t, err)
+	_, isJudge := client.(llmclient.Judge)
+	require.False(t, isJudge, "chat providers keep the Complete path")
+}
+
+func TestJevExtractorDoesNotSpendCallBudget(t *testing.T) {
+	t.Setenv("LLM_API_KEY", "key")
+	// ProxyFromEnvironment caches the first request in the process. Match the
+	// proxy the interactsh client test expects, and keep the local server direct.
+	for _, key := range []string{"HTTP_PROXY", "http_proxy", "ALL_PROXY", "all_proxy", "https_proxy", "no_proxy"} {
+		t.Setenv(key, "")
+	}
+	t.Setenv("HTTPS_PROXY", "http://127.0.0.1:9")
+	t.Setenv("NO_PROXY", "127.0.0.1,localhost")
+
+	var calls atomic.Int32
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		calls.Add(1)
+		_, _ = w.Write([]byte(`{"model":"jev-1.14.0","answers":{"q":{"type":"noul","noul":0.9}}}`))
+	}))
+	t.Cleanup(server.Close)
+
+	client, err := configureLLM(&types.Options{
+		EnableLLM:   true,
+		LLMProvider: "jev",
+		LLMBaseURL:  server.URL + "/v1",
+		LLMModel:    "jev-latest",
+		LLMMaxCalls: 1,
+	})
+	require.NoError(t, err)
+
+	_, err = client.Complete(context.Background(), "extract the title", true)
+	require.ErrorContains(t, err, "cannot run extractors")
+	_, err = client.Complete(context.Background(), "extract another", true)
+	require.Error(t, err)
+	require.Zero(t, calls.Load())
+
+	judge := client.(llmclient.Judge)
+	verdict, err := judge.Judge(context.Background(), "login page?", nil, []string{"<form>"})
+	require.NoError(t, err)
+	require.Equal(t, "yes", verdict.Label)
+	require.EqualValues(t, 1, calls.Load())
+}

internal/runner/options.go

@@ -228,25 +228,14 @@ func ValidateOptions(options *types.Options) error {
 		}
 	}
 
-	// verify that a valid ip version type was selected (4, 6)
-	if len(options.IPVersion) == 0 {
-		// add ipv4 as default
-		options.IPVersion = append(options.IPVersion, "4")
-	}
-	var useIPV4, useIPV6 bool
+	// An empty -iv leaves one unpinned target so the dialer can fall back
+	// between address families. -sa resolves and scans every selected address.
+	// Explicit -iv 4, -iv 6 and -iv 4,6 retain their usual selection semantics.
 	for _, ipv := range options.IPVersion {
-		switch ipv {
-		case "4":
-			useIPV4 = true
-		case "6":
-			useIPV6 = true
-		default:
+		if ipv != "4" && ipv != "6" {
 			return fmt.Errorf("unsupported ip version: %s", ipv)
 		}
 	}
-	if !useIPV4 && !useIPV6 {
-		return errors.New("ipv4 and/or ipv6 must be selected")
-	}
 	return nil
 }
 

internal/runner/options_test.go

@@ -90,3 +90,15 @@ func TestReadsTargetsFromStdin(t *testing.T) {
 		})
 	}
 }
+
+// The default must remain empty so the provider can distinguish automatic
+// IPv6 preference from explicit user-supplied -iv selections.
+func TestValidateOptionsIPVersionPreference(t *testing.T) {
+	automatic := &types.Options{}
+	require.NoError(t, ValidateOptions(automatic))
+	require.Empty(t, automatic.IPVersion)
+	for _, versions := range []goflags.StringSlice{{"4"}, {"6"}, {"4", "6"}} {
+		require.NoError(t, ValidateOptions(&types.Options{IPVersion: versions}))
+	}
+	require.ErrorContains(t, ValidateOptions(&types.Options{IPVersion: goflags.StringSlice{"7"}}), "unsupported ip version")
+}

internal/runner/runner.go

@@ -67,6 +67,7 @@ import (
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/utils/excludematchers"
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/headless/engine"
 	httpProtocol "github.com/projectdiscovery/nuclei/v3/pkg/protocols/http"
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/http/catchall"
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/http/httpclientpool"
 	"github.com/projectdiscovery/nuclei/v3/pkg/reporting"
 	"github.com/projectdiscovery/nuclei/v3/pkg/templates"
@@ -734,6 +735,9 @@ func (r *Runner) RunEnumeration() error {
 	if r.options.HostBackoff {
 		executorOpts.HostBackoff = hostbackoff.New(hostbackoff.Config{Max: r.options.HostBackoffMax})
 	}
+	if r.options.AutoCalibrate {
+		executorOpts.CatchAll = catchall.New(0)
+	}
 
 	executorEngine := core.New(r.options)
 	executorEngine.SetExecuterOptions(executorOpts)
@@ -1072,6 +1076,7 @@ func (r *Runner) displayExecutionInfo(store *loader.Store) {
 	stats.DisplayAsWarning(httpProtocol.SetThreadToCountZero)
 	stats.ForceDisplayWarning(templates.SkippedUnverifiedTemplateStats)
 	stats.ForceDisplayWarning(templates.SkippedRequestSignatureTemplateStats)
+	stats.ForceDisplayWarning(templates.SkippedWorkflowStepStats)
 
 	cfg := config.DefaultConfig
 

internal/runner/runner_test.go

@@ -6,11 +6,17 @@ import (
 	"net/http/httptest"
 	"os"
 	"path/filepath"
+	"strconv"
+	"sync/atomic"
 	"testing"
 	"time"
 
+	"github.com/miekg/dns"
+	"github.com/projectdiscovery/gologger"
 	"github.com/stretchr/testify/require"
 
+	"github.com/projectdiscovery/nuclei/v3/pkg/input"
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/provider"
 	"github.com/projectdiscovery/nuclei/v3/pkg/model/types/severity"
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/protocolstate"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
@@ -197,3 +203,63 @@ func TestNewInteractshHTTPClient(t *testing.T) {
 	require.NotNil(t, proxyURL)
 	require.Equal(t, "http://127.0.0.1:9", proxyURL.String())
 }
+
+func TestInitializeTemplatesHTTPInputUsesResolvers(t *testing.T) {
+	for _, key := range []string{"HTTP_PROXY", "http_proxy", "HTTPS_PROXY", "https_proxy"} {
+		t.Setenv(key, "")
+	}
+
+	listener, err := net.ListenPacket("udp", "127.0.0.1:0")
+	require.NoError(t, err)
+	queries := new(atomic.Int32)
+	server := &dns.Server{PacketConn: listener, Handler: dns.HandlerFunc(func(writer dns.ResponseWriter, request *dns.Msg) {
+		queries.Add(1)
+		response := new(dns.Msg).SetReply(request)
+		for _, question := range request.Question {
+			if question.Qtype == dns.TypeA {
+				response.Answer = append(response.Answer, &dns.A{Hdr: dns.RR_Header{Name: question.Name, Rrtype: dns.TypeA, Class: dns.ClassINET}, A: net.ParseIP("127.0.0.1")})
+			}
+		}
+		_ = writer.WriteMsg(response)
+	})}
+	started := make(chan struct{})
+	server.NotifyStartedFunc = func() { close(started) }
+	done := make(chan error, 1)
+	go func() { done <- server.ActivateAndServe() }()
+	<-started
+	t.Cleanup(func() { require.NoError(t, server.Shutdown()); require.NoError(t, <-done) })
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		w.WriteHeader(http.StatusOK)
+	}))
+	t.Cleanup(srv.Close)
+
+	resolverFile := filepath.Join(t.TempDir(), "resolv.conf")
+	require.NoError(t, os.WriteFile(resolverFile, nil, 0o600))
+	t.Setenv("RESOLVERS_PATH", resolverFile)
+
+	options := &types.Options{
+		ExecutionId:           t.Name(),
+		Timeout:               5,
+		BulkSize:              1,
+		InternalResolversList: []string{listener.LocalAddr().String()},
+	}
+	require.NoError(t, protocolstate.Init(options))
+	t.Cleanup(func() { protocolstate.Close(options.ExecutionId) })
+
+	target := "probe.nuclei.test:" + strconv.Itoa(srv.Listener.Addr().(*net.TCPAddr).Port)
+	r := &Runner{
+		options:       options,
+		inputProvider: provider.NewSimpleInputProviderWithUrls(options.ExecutionId, target),
+		Logger:        &gologger.Logger{},
+	}
+
+	hm, err := r.initializeTemplatesHTTPInput()
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = hm.Close() })
+
+	probed, ok := hm.Get(input.ProbeKey(target, ""))
+	require.True(t, ok, "the probe should resolve the target through the -r resolver")
+	require.Equal(t, "http://"+target, string(probed))
+	require.Positive(t, queries.Load())
+}

internal/runner/templates.go

@@ -10,9 +10,11 @@ import (
 	"github.com/logrusorgru/aurora/v4"
 	"github.com/projectdiscovery/nuclei/v3/pkg/catalog/config"
 	"github.com/projectdiscovery/nuclei/v3/pkg/catalog/loader"
+	"github.com/projectdiscovery/nuclei/v3/pkg/model"
 
 	"github.com/projectdiscovery/nuclei/v3/pkg/templates"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/projectdiscovery/nuclei/v3/pkg/utils"
 	"github.com/projectdiscovery/nuclei/v3/pkg/utils/json"
 )
 
@@ -38,12 +40,17 @@ func (r *Runner) verboseTemplate(tpl *templates.Template) {
 		tpl.Info.SeverityHolder.Severity))
 }
 
+// listedTemplate names its fields like a result event, so a -jsonl listing
+// joins with scan results on the same keys.
+type listedTemplate struct {
+	Template     string     `json:"template,omitempty"`
+	TemplateID   string     `json:"template-id"`
+	TemplatePath string     `json:"template-path"`
+	Info         model.Info `json:"info"`
+}
+
 func (r *Runner) listAvailableStoreTemplates(store *loader.Store) {
-	r.Logger.Print().Msgf(
-		"\nListing available %v nuclei templates for %v",
-		config.DefaultConfig.TemplateVersion,
-		config.DefaultConfig.TemplatesDirectory,
-	)
+	r.listingHeader("templates")
 	// order templates alphabetically by path
 	templates := store.Templates()
 	sort.Slice(templates, func(i, j int) bool {
@@ -51,6 +58,10 @@ func (r *Runner) listAvailableStoreTemplates(store *loader.Store) {
 	})
 
 	for _, tpl := range templates {
+		if r.options.JSONL && !r.options.TemplateDisplay {
+			r.listTemplateJSON(tpl)
+			continue
+		}
 		if hasExtraFlags(r.options) {
 			if r.options.TemplateDisplay {
 				colorize := !r.options.NoColor
@@ -78,12 +89,36 @@ func (r *Runner) listAvailableStoreTemplates(store *loader.Store) {
 	}
 }
 
-func (r *Runner) listAvailableTags(tagsMap map[string]int) {
+func (r *Runner) listTemplateJSON(tpl *templates.Template) {
+	path, _ := utils.TemplatePathURL(tpl.Path, tpl.ID, "")
+	marshalled, err := json.Marshal(listedTemplate{
+		Template:     path,
+		TemplateID:   tpl.ID,
+		TemplatePath: tpl.Path,
+		Info:         tpl.Info,
+	})
+	if err != nil {
+		r.Logger.Error().Msgf("Could not marshal the template %s: %s", tpl.Path, err)
+		return
+	}
+	r.Logger.Print().Msgf("%s", marshalled)
+}
+
+// listingHeader is skipped for -jsonl, where every line must parse as JSON.
+func (r *Runner) listingHeader(what string) {
+	if r.options.JSONL {
+		return
+	}
 	r.Logger.Print().Msgf(
-		"\nListing available %v nuclei tags for %v",
+		"\nListing available %v nuclei %s for %v",
 		config.DefaultConfig.TemplateVersion,
+		what,
 		config.DefaultConfig.TemplatesDirectory,
 	)
+}
+
+func (r *Runner) listAvailableTags(tagsMap map[string]int) {
+	r.listingHeader("tags")
 
 	type kv struct {
 		Key   string `json:"tag"`

internal/runner/templates_test.go

@@ -0,0 +1,76 @@
+package runner
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/projectdiscovery/gologger"
+	"github.com/projectdiscovery/gologger/formatter"
+	"github.com/projectdiscovery/gologger/levels"
+	"github.com/projectdiscovery/nuclei/v3/internal/tests/testutils"
+	"github.com/projectdiscovery/nuclei/v3/pkg/catalog/config"
+	"github.com/projectdiscovery/nuclei/v3/pkg/catalog/disk"
+	"github.com/projectdiscovery/nuclei/v3/pkg/catalog/loader"
+	"github.com/projectdiscovery/nuclei/v3/pkg/templates"
+	"github.com/projectdiscovery/nuclei/v3/pkg/utils/json"
+	"github.com/stretchr/testify/require"
+)
+
+const listedTemplateYAML = `id: listed-template
+info:
+  name: Listed Template
+  author: pdteam
+  severity: info
+  tags: listing
+http:
+  - method: GET
+    path:
+      - "{{BaseURL}}"
+`
+
+func TestTemplateListWritesJSONLines(t *testing.T) {
+	root := t.TempDir()
+	cfg := config.DefaultConfig
+	oldRoot := cfg.TemplatesDirectory
+	t.Cleanup(func() { cfg.SetTemplatesDir(oldRoot) })
+	cfg.SetTemplatesDir(root)
+
+	path := filepath.Join(root, "http", "listed-template.yaml")
+	require.NoError(t, os.MkdirAll(filepath.Dir(path), 0o755))
+	require.NoError(t, os.WriteFile(path, []byte(listedTemplateYAML), 0o600))
+
+	writer := &ignoreFileLogWriter{}
+	logger := &gologger.Logger{}
+	logger.SetFormatter(formatter.NewCLI(true))
+	logger.SetWriter(writer)
+	logger.SetMaxLevel(levels.LevelInfo)
+
+	options := testutils.DefaultOptions.Copy()
+	options.ExecutionId = t.Name()
+	options.Logger = logger
+	options.Templates = []string{path}
+	options.TemplateList = true
+	options.JSONL = true
+	testutils.Init(options)
+	t.Cleanup(func() { testutils.Cleanup(options) })
+
+	executerOpts := testutils.NewMockExecuterOptions(options, nil)
+	executerOpts.Parser = templates.NewParser()
+	store, err := loader.New(loader.NewConfig(options, disk.NewCatalog(root), executerOpts))
+	require.NoError(t, err)
+	require.NoError(t, store.LoadTemplatesOnlyMetadata())
+
+	require.Len(t, store.Templates(), 1)
+
+	r := &Runner{options: options, Logger: logger}
+	r.listAvailableStoreTemplates(store)
+
+	require.Len(t, writer.records, 1, "JSONL output carries no banner line")
+	var listed map[string]any
+	require.NoError(t, json.Unmarshal([]byte(strings.TrimSpace(writer.records[0].message)), &listed))
+	require.Equal(t, "listed-template", listed["template-id"])
+	require.Equal(t, filepath.Join("http", "listed-template.yaml"), listed["template"])
+	require.Equal(t, "Listed Template", listed["info"].(map[string]any)["name"])
+}

internal/server/nuclei_sdk.go

@@ -33,6 +33,7 @@ import (
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/interactsh"
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/utils/excludematchers"
 	browserEngine "github.com/projectdiscovery/nuclei/v3/pkg/protocols/headless/engine"
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/http/catchall"
 	"github.com/projectdiscovery/nuclei/v3/pkg/reporting"
 	"github.com/projectdiscovery/nuclei/v3/pkg/templates"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
@@ -110,6 +111,9 @@ func newNucleiExecutor(opts *NucleiExecutorOptions) (*nucleiExecutor, error) {
 	if opts.Options.HostBackoff {
 		executorOpts.HostBackoff = hostbackoff.New(hostbackoff.Config{Max: opts.Options.HostBackoffMax})
 	}
+	if opts.Options.AutoCalibrate {
+		executorOpts.CatchAll = catchall.New(0)
+	}
 
 	executorEngine := core.New(opts.Options)
 	executorEngine.SetExecuterOptions(executorOpts)

internal/tests/integration/http_test.go

@@ -102,6 +102,7 @@ var httpTestcases = []integrationCase{
 	{Path: "protocols/http/raw-unsafe-path-single-slash.yaml", TestCase: &httpRawUnsafePathSingleSlash{}},
 	{Path: "protocols/http/disable-http-cache.yaml", TestCase: &httpDisableCache{}},
 	{Path: "protocols/http/http-cache.yaml", TestCase: &httpCache{}},
+	{Path: "protocols/http/get.yaml", TestCase: &httpExcludeHostsFileComments{}},
 }
 
 type httpMultiVarSharing struct{}
@@ -1983,3 +1984,59 @@ func (h *httpDisableCache) Execute(filePath string) error {
 
 	return nil
 }
+
+// ensure that comments in an -exclude-hosts file are ignored
+type httpExcludeHostsFileComments struct{}
+
+// Execute executes a test case and returns an error if occurred
+func (h *httpExcludeHostsFileComments) Execute(filePath string) error {
+	router := httprouter.New()
+	router.GET("/", func(w http.ResponseWriter, r *http.Request, _ httprouter.Params) {
+		_, _ = fmt.Fprintf(w, "This is test matcher text")
+	})
+	ts := httptest.NewServer(router)
+	defer ts.Close()
+
+	excludeHosts := func(content string) ([]string, error) {
+		file, err := os.CreateTemp("", "exclude-hosts-*.txt")
+		if err != nil {
+			return nil, err
+		}
+		defer func() { _ = os.Remove(file.Name()) }()
+		_, err = file.WriteString(content)
+		_ = file.Close()
+		if err != nil {
+			return nil, err
+		}
+
+		return testutils.RunNucleiTemplateAndGetResults(filePath, ts.URL, debug, "-eh", file.Name())
+	}
+
+	// a comment line is not a host, even when it reads as an invalid regex
+	results, err := excludeHosts("# hosts out of scope (see ticket\n203.0.113.10\n")
+	if err != nil {
+		return fmt.Errorf("comment line: %w", err)
+	}
+	if err := expectResultsCount(results, 1); err != nil {
+		return fmt.Errorf("comment line: %w", err)
+	}
+
+	// a trailing comment does not stop the host from being excluded
+	results, err = excludeHosts("127.0.0.1 # test server\n")
+	if err != nil {
+		return fmt.Errorf("trailing comment: %w", err)
+	}
+	if err := expectResultsCount(results, 0); err != nil {
+		return fmt.Errorf("trailing comment: %w", err)
+	}
+
+	// comma separated hosts are still split when passed on the command line
+	results, err = testutils.RunNucleiTemplateAndGetResults(filePath, ts.URL, debug, "-eh", "203.0.113.10, 127.0.0.1")
+	if err != nil {
+		return fmt.Errorf("comma separated: %w", err)
+	}
+	if err := expectResultsCount(results, 0); err != nil {
+		return fmt.Errorf("comma separated: %w", err)
+	}
+	return nil
+}

lib/sdk_private.go

@@ -358,6 +358,9 @@ func (e *NucleiEngine) init(ctx context.Context) error {
 
 	httpxOptions := httpx.DefaultOptions
 	httpxOptions.Timeout = 5 * time.Second
+	if len(e.opts.InternalResolversList) > 0 {
+		httpxOptions.Resolvers = e.opts.InternalResolversList
+	}
 	if client, err := httpx.New(&httpxOptions); err != nil {
 		return err
 	} else {

pkg/catalog/loader/loader.go

@@ -631,10 +631,10 @@ func (store *Store) noteMissingCapabilities(templatePath string, missingCaps []t
 // erroring out on discovering any faulty templates.
 func (store *Store) ValidateTemplates() error {
 	templatePaths, errs := store.config.Catalog.GetTemplatesPath(store.finalTemplates)
-	store.logErroredTemplates(errs)
+	notFound := store.logErroredTemplates(errs)
 
 	workflowPaths, errs := store.config.Catalog.GetTemplatesPath(store.finalWorkflows)
-	store.logErroredTemplates(errs)
+	notFound += store.logErroredTemplates(errs)
 
 	templatePathsMap := make(map[string]struct{}, len(templatePaths))
 	for _, path := range templatePaths {
@@ -646,7 +646,8 @@ func (store *Store) ValidateTemplates() error {
 		workflowPathsMap[path] = struct{}{}
 	}
 
-	if store.areTemplatesValid(templatePathsMap) && store.areWorkflowsValid(workflowPathsMap) {
+	// a path that can't be resolved is not validated, so it can't count as valid
+	if store.areTemplatesValid(templatePathsMap) && store.areWorkflowsValid(workflowPathsMap) && notFound == 0 {
 		return nil
 	}
 
@@ -1056,10 +1057,15 @@ func workflowContainsProtocol(workflow []*workflows.WorkflowTemplate) bool {
 	return false
 }
 
-func (s *Store) logErroredTemplates(erred map[string]error) {
+// logErroredTemplates logs the templates that could not be found and returns
+// how many of them were not handled by the NotFoundCallback.
+func (s *Store) logErroredTemplates(erred map[string]error) int {
+	notFound := 0
 	for template, err := range erred {
 		if s.NotFoundCallback == nil || !s.NotFoundCallback(template) {
 			s.logger.Error().Msgf("Could not find template '%s': %s", template, err)
+			notFound++
 		}
 	}
+	return notFound
 }

pkg/catalog/loader/loader_test.go

@@ -176,25 +176,8 @@ func TestLoadTemplatesOnlyMetadataLogsCachedTemplateParseErrors(t *testing.T) {
 	require.Contains(t, output.String(), "Could not load template")
 }
 
-func TestValidateTemplatesRejectsRuntimeCompilationError(t *testing.T) {
-	templatePath := filepath.Join(t.TempDir(), "invalid-regex-group.yaml")
-	require.NoError(t, os.WriteFile(templatePath, []byte(`id: invalid-regex-group
-
-info:
-  name: Invalid Regex Group
-  author: pdteam
-  severity: info
-
-http:
-  - method: GET
-    path:
-      - "{{BaseURL}}"
-    extractors:
-      - type: regex
-        group: -1
-        regex:
-          - "(a)(b)"
-`), 0o600))
+func newValidateStore(t *testing.T, executionID, templatePath string) (*Store, *bytes.Buffer) {
+	t.Helper()
 
 	options := testutils.DefaultOptions.Copy()
 	var output bytes.Buffer
@@ -203,7 +186,7 @@ http:
 	logger.SetWriter(&utils.CaptureWriter{Buffer: &output})
 	logger.SetMaxLevel(levels.LevelDebug)
 	options.Logger = logger
-	options.ExecutionId = "loader-invalid-regex-group"
+	options.ExecutionId = executionID
 	options.Templates = []string{templatePath}
 	options.Validate = true
 	options.TemplateLoadingConcurrency = 1
@@ -226,10 +209,50 @@ http:
 
 	store, err := New(NewConfig(options, catalog, executerOpts))
 	require.NoError(t, err)
+	return store, &output
+}
+
+func TestValidateTemplatesRejectsRuntimeCompilationError(t *testing.T) {
+	templatePath := filepath.Join(t.TempDir(), "invalid-regex-group.yaml")
+	require.NoError(t, os.WriteFile(templatePath, []byte(`id: invalid-regex-group
+
+info:
+  name: Invalid Regex Group
+  author: pdteam
+  severity: info
+
+http:
+  - method: GET
+    path:
+      - "{{BaseURL}}"
+    extractors:
+      - type: regex
+        group: -1
+        regex:
+          - "(a)(b)"
+`), 0o600))
+
+	store, output := newValidateStore(t, "loader-invalid-regex-group", templatePath)
 	require.Error(t, store.ValidateTemplates())
 	require.Contains(t, output.String(), "regex extractor group must be >= 0")
 }
 
+func TestValidateTemplatesRejectsMissingTemplatePath(t *testing.T) {
+	templatePath := filepath.Join(t.TempDir(), "no-such-templates")
+
+	store, output := newValidateStore(t, "loader-missing-template-path", templatePath)
+	require.Error(t, store.ValidateTemplates())
+	require.Contains(t, output.String(), "Could not find template")
+}
+
+func TestValidateTemplatesAcceptsMissingPathHandledByCallback(t *testing.T) {
+	templatePath := filepath.Join(t.TempDir(), "no-such-templates")
+
+	store, _ := newValidateStore(t, "loader-missing-path-callback", templatePath)
+	store.NotFoundCallback = func(string) bool { return true }
+	require.NoError(t, store.ValidateTemplates())
+}
+
 func TestRemoteTemplates(t *testing.T) {
 	catalog := disk.NewCatalog("")
 

pkg/input/formats/README.md

@@ -5,8 +5,8 @@ Formats implements support for passing a number of request source as input provi
 Currently the following formats are implemented - 
 
 - Burp Suite XML Request/Response file
+- HAR (HTTP Archive) file exported by browsers and proxies
 - Proxify JSONL output file
-- Katana JSONL crawl output file
 - OpenAPI Specification file
 - Postman Collection file
 - Swagger Specification file
@@ -107,16 +107,3 @@ Swagger specification file is converted from OpenAPI 2.0 format to OpenAPI 3.0 f
 ## Burp XML / Proxify JSONL
 
 These modules are generic and parse raw requests from these respective tools.
-
-## Katana JSONL
-
-This module ingests the JSONL crawl output of the [katana](https://github.com/projectdiscovery/katana) crawler (`katana -jsonl -o crawl.jsonl`), bridging crawling and DAST so a live crawl of a target can directly drive fuzzing (`nuclei -im katana -l crawl.jsonl -dast ...`).
-
-For each crawled endpoint it produces a fuzzable request:
-
-- When katana captured the raw request (the common case, including authenticated and non-GET requests), the raw request is parsed as-is, preserving method, headers, body, and all parameters.
-- When only discrete fields are present (method, endpoint, headers, body), a raw request is synthesized from them and parsed back, so the result has the same shape as every other input format.
-
-The parser is tolerant of mixed input: blank and malformed lines are skipped with a warning, and a bare URL line (katana's default non-JSONL output) is treated as a `GET` request.
-
-A single record may be up to 32MB, which is well above a realistic worst case crawl record. Larger records, including a malformed one that never terminates, are drained and skipped with a warning rather than buffered, so one bad line cannot exhaust memory and end the scan.

pkg/input/formats/formats.go

@@ -12,8 +12,7 @@ import (
 	"gopkg.in/yaml.v3"
 )
 
-// ParseReqRespCallback is a callback function for discovered raw requests.
-// It returns false to ask the format to stop parsing.
+// ParseReqRespCallback is a callback function for discovered raw requests
 type ParseReqRespCallback func(rr *types.RequestResponse) bool
 
 // InputFormatOptions contains options for the input

pkg/input/formats/har/har.go

@@ -0,0 +1,137 @@
+package har
+
+import (
+	"encoding/base64"
+	"io"
+	"net/http"
+	"strconv"
+	"strings"
+
+	"github.com/pkg/errors"
+	"github.com/projectdiscovery/gologger"
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats"
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/types"
+	"github.com/projectdiscovery/nuclei/v3/pkg/utils/json"
+)
+
+// HARFormat is an HTTP Archive (HAR 1.2) parser, the format browsers and
+// proxies export captured traffic in.
+type HARFormat struct {
+	opts formats.InputFormatOptions
+}
+
+// New creates a new HAR parser
+func New() *HARFormat {
+	return &HARFormat{}
+}
+
+var _ formats.Format = &HARFormat{}
+
+type archive struct {
+	Log struct {
+		Entries []entry `json:"entries"`
+	} `json:"log"`
+}
+
+type entry struct {
+	Request struct {
+		Method   string   `json:"method"`
+		URL      string   `json:"url"`
+		Headers  []header `json:"headers"`
+		PostData *struct {
+			Text string `json:"text"`
+		} `json:"postData"`
+	} `json:"request"`
+	Response struct {
+		Status     int      `json:"status"`
+		StatusText string   `json:"statusText"`
+		Headers    []header `json:"headers"`
+		Content    struct {
+			Text     string `json:"text"`
+			Encoding string `json:"encoding"`
+		} `json:"content"`
+	} `json:"response"`
+}
+
+type header struct {
+	Name  string `json:"name"`
+	Value string `json:"value"`
+}
+
+// Name returns the name of the format
+func (h *HARFormat) Name() string {
+	return "har"
+}
+
+func (h *HARFormat) SetOptions(options formats.InputFormatOptions) {
+	h.opts = options
+}
+
+// Parse parses the input and calls the provided callback
+// function for each RawRequest it discovers.
+func (h *HARFormat) Parse(input io.Reader, resultsCb formats.ParseReqRespCallback, filePath string) error {
+	var har archive
+	if err := json.NewDecoder(input).Decode(&har); err != nil {
+		return errors.Wrap(err, "could not decode har file")
+	}
+
+	for _, e := range har.Log.Entries {
+		// Browsers also log data:, blob: and extension URLs, which have no
+		// host to send a request to.
+		if !strings.HasPrefix(e.Request.URL, "http://") && !strings.HasPrefix(e.Request.URL, "https://") {
+			continue
+		}
+		rawRequest, err := types.ParseRawRequestWithURL(rawRequestOf(e), e.Request.URL)
+		if err != nil {
+			gologger.Warning().Msgf("har: Could not parse request %s: %s\n", e.Request.URL, err)
+			continue
+		}
+		// Status 0 is a request that was blocked or never answered.
+		if e.Response.Status != 0 {
+			rawRequest.Response = &types.HttpResponse{Raw: rawResponseOf(e)}
+		}
+		resultsCb(rawRequest)
+	}
+	return nil
+}
+
+// rawRequestOf rebuilds the request as HTTP/1.1 with an absolute target, which
+// carries the authority that HTTP/2 captures only keep in pseudo-headers.
+func rawRequestOf(e entry) string {
+	var b strings.Builder
+	b.WriteString(e.Request.Method + " " + e.Request.URL + " HTTP/1.1\r\n")
+	writeHeaders(&b, e.Request.Headers)
+	if e.Request.PostData != nil {
+		b.WriteString(e.Request.PostData.Text)
+	}
+	return b.String()
+}
+
+func rawResponseOf(e entry) string {
+	statusText := e.Response.StatusText
+	if statusText == "" {
+		statusText = http.StatusText(e.Response.Status)
+	}
+	var b strings.Builder
+	b.WriteString("HTTP/1.1 " + strconv.Itoa(e.Response.Status) + " " + statusText + "\r\n")
+	writeHeaders(&b, e.Response.Headers)
+	body := e.Response.Content.Text
+	if e.Response.Content.Encoding == "base64" {
+		if decoded, err := base64.StdEncoding.DecodeString(body); err == nil {
+			body = string(decoded)
+		}
+	}
+	b.WriteString(body)
+	return b.String()
+}
+
+func writeHeaders(b *strings.Builder, headers []header) {
+	for _, h := range headers {
+		// HTTP/2 pseudo-headers (:authority, :path, ...) are not headers.
+		if strings.HasPrefix(h.Name, ":") {
+			continue
+		}
+		b.WriteString(h.Name + ": " + h.Value + "\r\n")
+	}
+	b.WriteString("\r\n")
+}

pkg/input/formats/har/har_test.go

@@ -0,0 +1,71 @@
+package har
+
+import (
+	"os"
+	"strings"
+	"testing"
+
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/types"
+	"github.com/stretchr/testify/require"
+)
+
+const chromeHAR = "../testdata/chrome.har"
+
+func parse(t *testing.T, input string) []*types.RequestResponse {
+	t.Helper()
+	file, err := os.Open(input)
+	require.NoError(t, err)
+	defer func() { _ = file.Close() }()
+
+	var got []*types.RequestResponse
+	require.NoError(t, New().Parse(file, func(rr *types.RequestResponse) bool {
+		got = append(got, rr)
+		return false
+	}, input))
+	return got
+}
+
+func TestHARParse(t *testing.T) {
+	got := parse(t, chromeHAR)
+
+	urls := make([]string, 0, len(got))
+	for _, rr := range got {
+		urls = append(urls, rr.URL.String())
+	}
+	require.Equal(t, []string{
+		"https://ginandjuice.shop/catalog?category=Juice",
+		"https://ginandjuice.shop/login",
+		"https://ginandjuice.shop/blocked.js",
+	}, urls, "non-http entries such as data: URLs are skipped")
+}
+
+func TestHARParseHTTP2Request(t *testing.T) {
+	rr := parse(t, chromeHAR)[0]
+
+	require.Equal(t, "GET", rr.Request.Method)
+	cookie, _ := rr.Request.Headers.Get("cookie")
+	require.Equal(t, "session=abc123", cookie)
+	for _, name := range rr.Request.Headers.GetKeys() {
+		require.False(t, strings.HasPrefix(name, ":"), "pseudo-header %s is not a real header", name)
+	}
+
+	req, err := rr.BuildRequest()
+	require.NoError(t, err)
+	require.Equal(t, "ginandjuice.shop", req.Host)
+}
+
+func TestHARParseBodyAndResponse(t *testing.T) {
+	got := parse(t, chromeHAR)
+
+	login := got[1]
+	require.Equal(t, "POST", login.Request.Method)
+	require.Equal(t, "username=carlos&password=hunter2", login.Request.Body)
+	require.NotNil(t, login.Response)
+	require.Contains(t, login.Response.Raw, "302 Found")
+	require.Contains(t, login.Response.Raw, "Location: /my-account")
+
+	require.NotNil(t, got[0].Response)
+	require.Contains(t, got[0].Response.Raw, "<h1>Juice</h1>", "base64 content is decoded")
+
+	require.Nil(t, got[2].Response, "a request that never got an answer has no response")
+}

pkg/input/formats/katana/katana.go

@@ -1,283 +0,0 @@
-// Package katana implements an input format that ingests the JSONL output of
-// the katana crawler (`katana -jsonl`) and turns each crawled endpoint into a
-// fuzzable nuclei request. This bridges crawling and DAST: instead of fuzzing
-// only a proxy-fed list of requests, nuclei can consume a crawl of the target
-// directly, preserving method, headers, body and parameters discovered on the
-// wire (including non-GET requests, which a bare URL list cannot express).
-package katana
-
-import (
-	"bufio"
-	"bytes"
-	"errors"
-	"fmt"
-	"io"
-	"net/http"
-	"net/url"
-	"sort"
-	"strings"
-
-	"github.com/projectdiscovery/gologger"
-	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats"
-	"github.com/projectdiscovery/nuclei/v3/pkg/input/types"
-	"github.com/projectdiscovery/nuclei/v3/pkg/utils/json"
-)
-
-const (
-	// MaxRecordSize is the maximum size of a single katana JSONL record. Crawled
-	// requests can carry large bodies and raw dumps, so the limit sits well above
-	// the 10MB of a realistic worst case record, while still bounding the memory a
-	// single record can consume. Records above it are skipped, not buffered.
-	MaxRecordSize = 32 * 1024 * 1024
-
-	// readBufferSize is the size of the chunks a record is read in. An oversized or
-	// unterminated record is drained through this buffer, so the memory used while
-	// skipping it stays constant regardless of how long the record is.
-	readBufferSize = 64 * 1024
-)
-
-// KatanaFormat is a parser for katana JSONL crawl output.
-type KatanaFormat struct {
-	opts formats.InputFormatOptions
-}
-
-// New creates a new katana JSONL format parser.
-func New() *KatanaFormat {
-	return &KatanaFormat{}
-}
-
-var _ formats.Format = &KatanaFormat{}
-
-// katanaResult mirrors the relevant subset of katana's output.Result JSON.
-type katanaResult struct {
-	Request *katanaRequest `json:"request"`
-	Error   string         `json:"error"`
-}
-
-// katanaRequest mirrors the relevant subset of katana's navigation.Request JSON.
-type katanaRequest struct {
-	Method   string            `json:"method"`
-	Endpoint string            `json:"endpoint"`
-	Body     string            `json:"body"`
-	Headers  map[string]string `json:"headers"`
-	Raw      string            `json:"raw"`
-}
-
-// Name returns the name of the format.
-func (k *KatanaFormat) Name() string {
-	return "katana"
-}
-
-// SetOptions sets the options for the input format.
-func (k *KatanaFormat) SetOptions(options formats.InputFormatOptions) {
-	k.opts = options
-}
-
-// Parse parses katana JSONL output and calls the provided callback for each
-// request it discovers. It is tolerant of mixed input: blank lines are skipped,
-// and a bare URL line (katana's default non-JSONL output) is treated as a GET.
-// Records above MaxRecordSize are dropped with a warning so that a malformed or
-// unterminated record cannot exhaust memory.
-func (k *KatanaFormat) Parse(input io.Reader, resultsCb formats.ParseReqRespCallback, filePath string) error {
-	reader := bufio.NewReaderSize(input, readBufferSize)
-	for {
-		record, oversized, err := readRecord(reader)
-		if err != nil && !errors.Is(err, io.EOF) {
-			return fmt.Errorf("could not read katana jsonl input: %w", err)
-		}
-
-		if oversized {
-			gologger.Warning().Msgf("katana: skipping record larger than %d bytes\n", MaxRecordSize)
-		} else if len(record) > 0 && !k.handleRecord(record, resultsCb) {
-			return nil
-		}
-
-		if errors.Is(err, io.EOF) {
-			return nil
-		}
-	}
-}
-
-// readRecord reads a single newline delimited record, keeping at most
-// MaxRecordSize bytes in memory. A record above the limit is drained chunk by
-// chunk and reported as oversized so the caller can skip it, instead of being
-// accumulated in full.
-func readRecord(reader *bufio.Reader) (record []byte, oversized bool, err error) {
-	for {
-		chunk, err := reader.ReadSlice('\n')
-		if len(chunk) > 0 {
-			// The delimiter is trimmed off the record, so it does not count
-			// towards the limit.
-			size := len(chunk)
-			if chunk[size-1] == '\n' {
-				size--
-			}
-			if oversized || len(record)+size > MaxRecordSize {
-				// Release what was read so far and keep consuming until the record
-				// ends: only the reader buffer stays allocated while draining.
-				oversized, record = true, nil
-			} else {
-				record = append(record, chunk...)
-			}
-		}
-		// A full buffer without a delimiter means the record continues.
-		if errors.Is(err, bufio.ErrBufferFull) {
-			continue
-		}
-		return bytes.TrimSpace(record), oversized, err
-	}
-}
-
-// handleRecord parses a single record and reports whether parsing should
-// continue. A skipped record continues; the callback returns false to stop.
-func (k *KatanaFormat) handleRecord(record []byte, resultsCb formats.ParseReqRespCallback) bool {
-	if record[0] != '{' {
-		// Bare URL line (katana default output without -jsonl): treat as GET.
-		line := string(record)
-		if !isAbsoluteURL(line) {
-			gologger.Warning().Msg("katana: could not parse line as a URL or JSON record\n")
-			return true
-		}
-		rr, err := k.buildFromComponents(http.MethodGet, line, nil, "")
-		if err != nil {
-			gologger.Warning().Msgf("katana: could not parse url %s: %s\n", line, err)
-			return true
-		}
-		return resultsCb(rr)
-	}
-
-	var result katanaResult
-	if err := json.Unmarshal(record, &result); err != nil {
-		gologger.Warning().Msgf("katana: could not decode jsonl line: %s\n", err)
-		return true
-	}
-	if result.Request == nil || result.Request.Endpoint == "" {
-		gologger.Warning().Msg("katana: invalid record with missing request or endpoint\n")
-		return true
-	}
-	if !isAbsoluteURL(result.Request.Endpoint) {
-		gologger.Warning().Msgf("katana: skipping non http(s) endpoint %s\n", result.Request.Endpoint)
-		return true
-	}
-	rr, err := k.toRequestResponse(result.Request)
-	if err != nil {
-		gologger.Warning().Msgf("katana: could not parse request %s: %s\n", result.Request.Endpoint, err)
-		return true
-	}
-	return resultsCb(rr)
-}
-
-// toRequestResponse converts a katana request into nuclei's standard
-// RequestResponse. It prefers the captured raw request when available and falls
-// back to synthesizing one from the discovered components otherwise.
-func (k *KatanaFormat) toRequestResponse(req *katanaRequest) (*types.RequestResponse, error) {
-	method := req.Method
-	if method == "" {
-		method = http.MethodGet
-	}
-	if strings.TrimSpace(req.Raw) == "" {
-		return k.buildFromComponents(method, req.Endpoint, req.Headers, req.Body)
-	}
-	rr, err := types.ParseRawRequestWithURL(req.Raw, req.Endpoint)
-	if err != nil || !rawConflicts(req, rr) {
-		return rr, err
-	}
-	// Some katana versions record a form submission as a POST with a body but
-	// keep a raw GET without it (katana issue 876). Rebuild the request from the
-	// structured fields so the method and body survive, and keep the headers the
-	// raw request captured, such as cookies.
-	return k.buildFromComponents(method, req.Endpoint, mergeRawHeaders(rr, req.Headers), req.Body)
-}
-
-// rawConflicts reports whether the parsed raw request disagrees with the
-// structured method or body of the katana request.
-func rawConflicts(req *katanaRequest, rr *types.RequestResponse) bool {
-	if rr == nil || rr.Request == nil {
-		return false
-	}
-	if req.Method != "" && !strings.EqualFold(req.Method, rr.Request.Method) {
-		return true
-	}
-	return req.Body != "" && rr.Request.Body == ""
-}
-
-// mergeRawHeaders returns the headers of the raw request overridden by the
-// structured headers. Framing headers are dropped, because they describe the
-// raw body and not the structured one.
-func mergeRawHeaders(rr *types.RequestResponse, structured map[string]string) map[string]string {
-	headers := make(map[string]string)
-	rr.Request.Headers.Iterate(func(key, value string) bool {
-		if !strings.EqualFold(key, "Content-Length") && !strings.EqualFold(key, "Transfer-Encoding") {
-			headers[http.CanonicalHeaderKey(key)] = value
-		}
-		return true
-	})
-	for key, value := range structured {
-		headers[http.CanonicalHeaderKey(key)] = value
-	}
-	return headers
-}
-
-// buildFromComponents synthesizes a raw HTTP request from discrete fields and
-// parses it back into a RequestResponse, reusing the well-tested raw parser so
-// the resulting object is identical in shape to the other input formats.
-func (k *KatanaFormat) buildFromComponents(method, endpoint string, headers map[string]string, body string) (*types.RequestResponse, error) {
-	parsed, err := url.Parse(endpoint)
-	if err != nil {
-		return nil, fmt.Errorf("invalid endpoint: %w", err)
-	}
-	if parsed.Host == "" {
-		return nil, fmt.Errorf("endpoint %q has no host", endpoint)
-	}
-
-	requestURI := parsed.RequestURI()
-	if requestURI == "" {
-		requestURI = "/"
-	}
-
-	var sb strings.Builder
-	sb.WriteString(method)
-	sb.WriteString(" ")
-	sb.WriteString(requestURI)
-	sb.WriteString(" HTTP/1.1\r\n")
-	sb.WriteString("Host: ")
-	sb.WriteString(parsed.Host)
-	sb.WriteString("\r\n")
-
-	// Emit headers in a deterministic order, skipping Host (already written)
-	// and any header with a line break, which would split the raw request.
-	for _, key := range sortedHeaderKeys(headers) {
-		if strings.EqualFold(key, "Host") || strings.ContainsAny(key+headers[key], "\r\n") {
-			continue
-		}
-		sb.WriteString(key)
-		sb.WriteString(": ")
-		sb.WriteString(headers[key])
-		sb.WriteString("\r\n")
-	}
-	sb.WriteString("\r\n")
-	if body != "" {
-		sb.WriteString(body)
-	}
-
-	return types.ParseRawRequestWithURL(sb.String(), endpoint)
-}
-
-// sortedHeaderKeys returns the header keys in deterministic (sorted) order.
-func sortedHeaderKeys(headers map[string]string) []string {
-	if len(headers) == 0 {
-		return nil
-	}
-	keys := make([]string, 0, len(headers))
-	for k := range headers {
-		keys = append(keys, k)
-	}
-	sort.Strings(keys)
-	return keys
-}
-
-// isAbsoluteURL reports whether the line is an absolute http(s) URL.
-func isAbsoluteURL(line string) bool {
-	u, err := url.Parse(line)
-	return err == nil && u.IsAbs() && u.Host != "" && (u.Scheme == "http" || u.Scheme == "https")
-}

pkg/input/formats/katana/katana_test.go

@@ -1,245 +0,0 @@
-package katana
-
-import (
-	"io"
-	"os"
-	"runtime"
-	"strings"
-	"testing"
-
-	"github.com/projectdiscovery/nuclei/v3/pkg/input/types"
-	"github.com/stretchr/testify/require"
-)
-
-func TestKatanaFormatName(t *testing.T) {
-	require.Equal(t, "katana", New().Name())
-}
-
-func TestKatanaFormatParse(t *testing.T) {
-	inputFile := "../testdata/katana.jsonl"
-
-	file, err := os.Open(inputFile)
-	require.Nilf(t, err, "error opening katana input file: %v", err)
-	defer func() { _ = file.Close() }()
-
-	var got []*types.RequestResponse
-	err = New().Parse(file, func(rr *types.RequestResponse) bool {
-		got = append(got, rr)
-		return true
-	}, inputFile)
-	require.NoError(t, err)
-
-	// 3 JSONL records + 1 bare URL line; malformed and blank lines skipped.
-	require.Len(t, got, 4)
-
-	urls := make([]string, 0, len(got))
-	for _, rr := range got {
-		urls = append(urls, rr.URL.String())
-	}
-	require.ElementsMatch(t, []string{
-		"https://ginandjuice.shop/catalog/product?productId=1",
-		"https://ginandjuice.shop/catalog/subscribe",
-		"https://ginandjuice.shop/login",
-		"https://ginandjuice.shop/about",
-	}, urls)
-}
-
-func TestKatanaFormatRawRequestPreserved(t *testing.T) {
-	// A POST with a captured raw request must preserve method and body.
-	rr := parseSingle(t, `{"request":{"method":"POST","endpoint":"https://example.com/sub","raw":"POST /sub HTTP/1.1\r\nHost: example.com\r\nContent-Type: application/x-www-form-urlencoded\r\n\r\nemail=a%40b.com"}}`)
-	require.Equal(t, "POST", rr.Request.Method)
-	require.Equal(t, "email=a%40b.com", rr.Request.Body)
-	require.Equal(t, "https://example.com/sub", rr.URL.String())
-}
-
-func TestKatanaFormatRawConflictUsesStructuredRequest(t *testing.T) {
-	// katana issue 876: a POST with a body whose raw request is a GET without it.
-	rr := parseSingle(t, `{"request":{"method":"POST","endpoint":"https://example.com/sub","body":"user=a&pass=b","headers":{"Content-Type":"application/x-www-form-urlencoded"},"raw":"GET /sub HTTP/1.1\r\nHost: example.com\r\nCookie: s=1\r\nContent-Length: 0\r\n\r\n"}}`)
-	require.Equal(t, "POST", rr.Request.Method)
-	require.Equal(t, "user=a&pass=b", rr.Request.Body)
-	require.Equal(t, "https://example.com/sub", rr.URL.String())
-
-	built, err := rr.BuildRequest()
-	require.NoError(t, err)
-	require.Equal(t, "s=1", built.Header.Get("Cookie"))
-	require.Equal(t, "application/x-www-form-urlencoded", built.Header.Get("Content-Type"))
-	require.Equal(t, int64(len("user=a&pass=b")), built.ContentLength)
-}
-
-func TestKatanaFormatComponentSynthesis(t *testing.T) {
-	// No raw request: it must be synthesized from method/endpoint/headers/body.
-	rr := parseSingle(t, `{"request":{"method":"POST","endpoint":"https://example.com/login","headers":{"Content-Type":"application/json"},"body":"{\"u\":\"admin\"}"}}`)
-	require.Equal(t, "POST", rr.Request.Method)
-	require.Equal(t, `{"u":"admin"}`, rr.Request.Body)
-	require.Equal(t, "https://example.com/login", rr.URL.String())
-
-	built, err := rr.BuildRequest()
-	require.NoError(t, err)
-	require.Equal(t, "application/json", built.Header.Get("Content-Type"))
-	require.Equal(t, "example.com", built.Host)
-}
-
-func TestKatanaFormatQueryParamsRetained(t *testing.T) {
-	// Parameter variants must survive so the fuzzer can target them.
-	rr := parseSingle(t, `{"request":{"method":"GET","endpoint":"https://example.com/p?id=1&q=2"}}`)
-	require.Equal(t, "https://example.com/p?id=1&q=2", rr.URL.String())
-	require.Equal(t, "GET", rr.Request.Method)
-}
-
-func TestKatanaFormatGracefulSkips(t *testing.T) {
-	input := strings.Join([]string{
-		"",
-		"garbage",
-		`{"request":null}`,
-		`{"request":{"endpoint":""}}`,
-		`{"request":{"method":"GET","endpoint":"ftp://example.com/file"}}`,
-		`{"request":{"method":"GET","endpoint":"/relative/only"}}`,
-		`{"request":{"method":"GET","endpoint":"https://example.com/ok"}}`,
-	}, "\n")
-
-	var got []*types.RequestResponse
-	err := New().Parse(strings.NewReader(input), func(rr *types.RequestResponse) bool {
-		got = append(got, rr)
-		return true
-	}, "test")
-	require.NoError(t, err)
-	require.Len(t, got, 1)
-	require.Equal(t, "https://example.com/ok", got[0].URL.String())
-}
-
-func TestKatanaFormatCallbackStops(t *testing.T) {
-	input := strings.Join([]string{
-		`{"request":{"method":"GET","endpoint":"https://example.com/1"}}`,
-		`{"request":{"method":"GET","endpoint":"https://example.com/2"}}`,
-	}, "\n")
-
-	var count int
-	err := New().Parse(strings.NewReader(input), func(_ *types.RequestResponse) bool {
-		count++
-		return false // request to stop after the first
-	}, "test")
-	require.NoError(t, err)
-	require.Equal(t, 1, count)
-}
-
-func TestKatanaFormatLargeRecord(t *testing.T) {
-	largeBody := strings.Repeat("a", 10*1024*1024+1)
-	input := `{"request":{"method":"POST","endpoint":"https://example.com/upload","body":"` + largeBody + `"}}`
-
-	rr := parseSingle(t, input)
-	require.Equal(t, largeBody, rr.Request.Body)
-}
-
-func TestKatanaFormatRecordSizeLimit(t *testing.T) {
-	const (
-		prefix = `{"request":{"method":"POST","endpoint":"https://example.com/upload","body":"`
-		suffix = `"}}`
-	)
-	record := func(size int) string {
-		return prefix + strings.Repeat("a", size-len(prefix)-len(suffix)) + suffix
-	}
-
-	// A record right at the limit is parsed, one byte over it is dropped.
-	rr := parseSingle(t, record(MaxRecordSize))
-	require.Len(t, rr.Request.Body, MaxRecordSize-len(prefix)-len(suffix))
-
-	var got []*types.RequestResponse
-	err := New().Parse(strings.NewReader(record(MaxRecordSize+1)), func(rr *types.RequestResponse) bool {
-		got = append(got, rr)
-		return true
-	}, "test")
-	require.NoError(t, err)
-	require.Empty(t, got)
-}
-
-func TestKatanaFormatOversizedRecordSkipped(t *testing.T) {
-	// A record above the limit must be dropped while the rest of the file is
-	// still parsed.
-	input := io.MultiReader(
-		strings.NewReader(`{"request":{"method":"POST","endpoint":"https://example.com/upload","body":"`),
-		io.LimitReader(filler{}, MaxRecordSize),
-		strings.NewReader("\"}}\n"+`{"request":{"method":"GET","endpoint":"https://example.com/ok"}}`),
-	)
-
-	var got []*types.RequestResponse
-	err := New().Parse(input, func(rr *types.RequestResponse) bool {
-		got = append(got, rr)
-		return true
-	}, "test")
-	require.NoError(t, err)
-	require.Len(t, got, 1)
-	require.Equal(t, "https://example.com/ok", got[0].URL.String())
-}
-
-func TestKatanaFormatUnterminatedRecordIsBounded(t *testing.T) {
-	// An unterminated record is drained rather than buffered, so what parsing
-	// costs must not grow with how long the record is: a record sixteen times
-	// over the limit may not cost measurably more than one just over it.
-	small := allocsForUnterminatedRecord(t, 2*MaxRecordSize)
-	large := allocsForUnterminatedRecord(t, 16*MaxRecordSize)
-
-	require.LessOrEqual(t, large, small+MaxRecordSize,
-		"draining a record 16x over the limit allocated %d bytes against %d for 2x", large, small)
-}
-
-// allocsForUnterminatedRecord reports the bytes allocated while parsing a
-// record of the given size that never terminates with a newline.
-func allocsForUnterminatedRecord(t *testing.T, size int64) uint64 {
-	t.Helper()
-
-	input := io.MultiReader(
-		strings.NewReader(`{"request":{"method":"POST","endpoint":"https://example.com/upload","body":"`),
-		io.LimitReader(filler{}, size),
-	)
-
-	var before, after runtime.MemStats
-	runtime.GC()
-	runtime.ReadMemStats(&before)
-
-	var got []*types.RequestResponse
-	err := New().Parse(input, func(rr *types.RequestResponse) bool {
-		got = append(got, rr)
-		return true
-	}, "test")
-	runtime.ReadMemStats(&after)
-
-	require.NoError(t, err)
-	require.Empty(t, got)
-
-	return after.TotalAlloc - before.TotalAlloc
-}
-
-// filler is an endless source of record payload bytes, used to feed oversized
-// records without materializing them in the test.
-type filler struct{}
-
-func (filler) Read(p []byte) (int, error) {
-	for i := range p {
-		p[i] = 'a'
-	}
-	return len(p), nil
-}
-
-func parseSingle(t *testing.T, line string) *types.RequestResponse {
-	t.Helper()
-	var got *types.RequestResponse
-	err := New().Parse(strings.NewReader(line), func(rr *types.RequestResponse) bool {
-		got = rr
-		return true
-	}, "test")
-	require.NoError(t, err)
-	require.NotNil(t, got)
-	return got
-}
-
-func TestKatanaFormatHeaderLineBreaksDropped(t *testing.T) {
-	// A crawled header value with a line break must not inject extra headers.
-	rr := parseSingle(t, `{"request":{"method":"GET","endpoint":"https://example.com/x","headers":{"X-Ok":"1","X-Bad":"a\r\nX-Injected: 1"}}}`)
-	value, ok := rr.Request.Headers.Get("X-Ok")
-	require.True(t, ok)
-	require.Equal(t, "1", value)
-	_, ok = rr.Request.Headers.Get("X-Injected")
-	require.False(t, ok)
-	_, ok = rr.Request.Headers.Get("X-Bad")
-	require.False(t, ok)
-}

pkg/input/formats/openapi/fuzz.go

@@ -0,0 +1,20 @@
+//go:build gofuzz
+// +build gofuzz
+
+package openapi
+
+// Fuzz exercises OpenAPI 3.0 schema parsing and request generation. The input
+// bytes are treated as a raw schema document (YAML or JSON) and fed through the
+// format's Parse, which loads the schema and generates requests from it.
+func Fuzz(data []byte) int {
+	if len(data) == 0 {
+		return 0
+	}
+	if len(data) > fuzzMaxInputSize {
+		return -1
+	}
+	if !fuzzOpenAPIParsing(data) {
+		return 0
+	}
+	return 1
+}

pkg/input/formats/openapi/fuzz_harness.go

@@ -0,0 +1,32 @@
+package openapi
+
+import (
+	"bytes"
+
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats"
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/types"
+)
+
+const fuzzMaxInputSize = 64 << 10
+
+// fuzzOpenAPIParsing feeds data to the OpenAPI format parser as a raw schema
+// document and drives request generation. It must never panic for any input;
+// it returns true when the schema parsed and at least one request was produced.
+func fuzzOpenAPIParsing(data []byte) bool {
+	if len(data) == 0 || len(data) > fuzzMaxInputSize {
+		return false
+	}
+
+	format := New()
+	format.SetOptions(formats.InputFormatOptions{})
+
+	produced := false
+	cb := func(rr *types.RequestResponse) bool {
+		produced = true
+		return true
+	}
+	if err := format.Parse(bytes.NewReader(data), cb, "fuzz.openapi"); err != nil {
+		return false
+	}
+	return produced
+}

pkg/input/formats/openapi/fuzz_harness_test.go

@@ -0,0 +1,84 @@
+package openapi
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/types"
+	"github.com/stretchr/testify/require"
+)
+
+func TestFuzzOpenAPIParsingSeedCorpus(t *testing.T) {
+	entries, err := os.ReadDir("testdata/gofuzz-corpus")
+	require.NoError(t, err)
+	require.NotEmpty(t, entries)
+
+	for _, entry := range entries {
+		if entry.IsDir() {
+			continue
+		}
+
+		data, err := os.ReadFile(filepath.Join("testdata/gofuzz-corpus", entry.Name()))
+		require.NoError(t, err)
+		require.Truef(t, fuzzOpenAPIParsing(data), "seed %s should produce at least one request", entry.Name())
+	}
+}
+
+func TestFuzzOpenAPIParsingRejectsOversizeInput(t *testing.T) {
+	require.False(t, fuzzOpenAPIParsing(make([]byte, fuzzMaxInputSize+1)))
+}
+
+// Null entries decode to nil pointers and used to panic during request
+// generation. The remaining valid entries must still produce requests.
+func TestFuzzOpenAPIParsingNullEntries(t *testing.T) {
+	specs := map[string]string{
+		"null server": `
+openapi: 3.0.0
+info: {title: t, version: "1"}
+servers:
+  -
+  - {url: "http://example.com"}
+paths:
+  /users:
+    get:
+      responses: {"200": {description: ok}}
+`,
+		"null media type": `
+openapi: 3.0.0
+info: {title: t, version: "1"}
+servers: [{url: "http://example.com"}]
+paths:
+  /users:
+    post:
+      requestBody:
+        content:
+          application/xml:
+          application/json:
+            schema: {type: object, properties: {name: {type: string}}}
+      responses: {"200": {description: ok}}
+`,
+	}
+
+	for name, spec := range specs {
+		t.Run(name, func(t *testing.T) {
+			require.True(t, fuzzOpenAPIParsing([]byte(spec)))
+		})
+	}
+}
+
+func TestGenerateRequestsOnlyNullServers(t *testing.T) {
+	spec := `
+openapi: 3.0.0
+info: {title: t, version: "1"}
+servers:
+  -
+paths:
+  /users:
+    get:
+      responses: {"200": {description: ok}}
+`
+	err := New().Parse(strings.NewReader(spec), func(*types.RequestResponse) bool { return false }, "spec.yaml")
+	require.ErrorContains(t, err, "no servers found in openapi schema")
+}

pkg/input/formats/openapi/generator.go

@@ -9,6 +9,7 @@ import (
 	"net/http/httputil"
 	"net/url"
 	"os"
+	"slices"
 	"strings"
 
 	"github.com/clbanning/mxj/v2"
@@ -33,7 +34,9 @@ const (
 
 // GenerateRequestsFromSchema generates http requests from an OpenAPI 3.0 document object
 func GenerateRequestsFromSchema(schema *openapi3.T, opts formats.InputFormatOptions, callback formats.ParseReqRespCallback) error {
-	if len(schema.Servers) == 0 {
+	// Null entries in `servers:` are skipped below, so a list of only null
+	// entries has no usable server, the same as an empty list.
+	if !slices.ContainsFunc(schema.Servers, func(s *openapi3.Server) bool { return s != nil }) {
 		return errors.New("no servers found in openapi schema")
 	}
 	globalParams, globalSecurityErr := selectSecurityParameters(schema, &schema.Security, opts.Variables)
@@ -49,6 +52,11 @@ func GenerateRequestsFromSchema(schema *openapi3.T, opts formats.InputFormatOpti
 	}
 
 	for _, serverURL := range schema.Servers {
+		// A null entry in the `servers:` list decodes to a nil *Server, so skip
+		// it rather than dereferencing it.
+		if serverURL == nil {
+			continue
+		}
 		pathURL := serverURL.URL
 		// Split the server URL into baseURL and serverPath
 		u, err := url.Parse(pathURL)
@@ -268,6 +276,11 @@ func generateRequestsFromOp(opts *generateReqOptions) error {
 
 	if opts.op.RequestBody != nil {
 		for content, value := range opts.op.RequestBody.Value.Content {
+			// A null media-type entry (e.g. `content: {application/json: }`)
+			// decodes to a nil *MediaType, so skip it before dereferencing.
+			if value == nil {
+				continue
+			}
 			cloned := req.Clone(req.Context())
 
 			var val interface{}

pkg/input/formats/openapi/testdata/gofuzz-corpus/json-spec.seed

@@ -0,0 +1 @@
+{"openapi":"3.0.0","info":{"title":"t","version":"1"},"servers":[{"url":"http://example.com"}],"paths":{"/a":{"post":{"requestBody":{"content":{"application/json":{"schema":{"type":"object","properties":{"k":{"type":"string"}}}}}},"responses":{"200":{"description":"ok"}}}}}}

pkg/input/formats/openapi/testdata/gofuzz-corpus/minimal.seed

@@ -0,0 +1,7 @@
+openapi: 3.0.0
+info: {title: t, version: "1"}
+servers: [{url: "http://example.com"}]
+paths:
+  /users:
+    get:
+      responses: {"200": {description: ok}}

pkg/input/formats/openapi/testdata/gofuzz-corpus/nil-media-type.seed

@@ -0,0 +1,12 @@
+openapi: 3.0.0
+info: {title: t, version: "1"}
+servers: [{url: "http://example.com"}]
+paths:
+  /users:
+    post:
+      requestBody:
+        content:
+          application/xml:
+          application/json:
+            schema: {type: object, properties: {name: {type: string}}}
+      responses: {"200": {description: ok}}

pkg/input/formats/openapi/testdata/gofuzz-corpus/nil-server.seed

@@ -0,0 +1,9 @@
+openapi: 3.0.0
+info: {title: t, version: "1"}
+servers:
+  -
+  - {url: "http://example.com"}
+paths:
+  /users:
+    get:
+      responses: {"200": {description: ok}}

pkg/input/formats/swagger/swagger.go

@@ -3,7 +3,9 @@ package swagger
 import (
 	"fmt"
 	"io"
+	"maps"
 	"path"
+	"slices"
 
 	"github.com/getkin/kin-openapi/openapi2"
 	"github.com/getkin/kin-openapi/openapi2conv"
@@ -55,6 +57,7 @@ func (j *SwaggerFormat) Parse(input io.Reader, resultsCb formats.ParseReqRespCal
 	if err != nil {
 		return errors.Wrap(err, "could not decode openapi 2.0 schema")
 	}
+	dropNullEntries(schemav2)
 	schema, err := openapi2conv.ToV3(schemav2)
 	if err != nil {
 		return errors.Wrap(err, "could not convert openapi 2.0 schema to 3.0")
@@ -67,6 +70,32 @@ func (j *SwaggerFormat) Parse(input io.Reader, resultsCb formats.ParseReqRespCal
 	return openapi.GenerateRequestsFromSchema(schema, j.opts, resultsCb)
 }
 
+// dropNullEntries removes null entries from the maps and lists of a decoded
+// OpenAPI 2.0 document. A null value such as `paths: {/x: }` or an empty list
+// item under `parameters:` decodes to a nil pointer, which openapi2conv.ToV3
+// dereferences without checking.
+func dropNullEntries(doc *openapi2.T) {
+	maps.DeleteFunc(doc.Paths, func(_ string, v *openapi2.PathItem) bool { return v == nil })
+	maps.DeleteFunc(doc.Parameters, func(_ string, v *openapi2.Parameter) bool { return v == nil })
+	dropNullResponses(doc.Responses)
+	for _, pathItem := range doc.Paths {
+		pathItem.Parameters = slices.DeleteFunc(pathItem.Parameters, func(v *openapi2.Parameter) bool { return v == nil })
+		for _, op := range pathItem.Operations() {
+			op.Parameters = slices.DeleteFunc(op.Parameters, func(v *openapi2.Parameter) bool { return v == nil })
+			dropNullResponses(op.Responses)
+		}
+	}
+}
+
+// dropNullResponses removes null responses, and null headers of the remaining
+// responses, which openapi2conv.ToV3Response also dereferences.
+func dropNullResponses(responses map[string]*openapi2.Response) {
+	maps.DeleteFunc(responses, func(_ string, v *openapi2.Response) bool { return v == nil })
+	for _, response := range responses {
+		maps.DeleteFunc(response.Headers, func(_ string, v *openapi2.Header) bool { return v == nil })
+	}
+}
+
 func decodeYAML(data []byte, target interface{}) error {
 	var value interface{}
 	if err := yaml.Unmarshal(data, &value); err != nil {

pkg/input/formats/swagger/swagger_test.go

@@ -2,6 +2,7 @@ package swagger
 
 import (
 	"os"
+	"strings"
 	"testing"
 
 	"github.com/projectdiscovery/nuclei/v3/pkg/input/types"
@@ -39,3 +40,69 @@ func TestSwaggerAPIParser(t *testing.T) {
 	}
 	require.ElementsMatch(t, gotMethodsToURLs, expectedURLs, "could not get swagger urls")
 }
+
+func TestSwaggerParseNullEntries(t *testing.T) {
+	// Each spec has a null map value or list item that decodes to a nil
+	// pointer and used to panic inside openapi2conv.ToV3.
+	specs := map[string]string{
+		"null-path-item": `
+swagger: "2.0"
+info: {title: t, version: "1"}
+host: localhost
+paths:
+  /x:
+`,
+		"null-parameter": `
+swagger: "2.0"
+info: {title: t, version: "1"}
+host: localhost
+paths:
+  /x:
+    get:
+      parameters:
+        -
+      responses: {"200": {description: ok}}
+`,
+		"null-response": `
+swagger: "2.0"
+info: {title: t, version: "1"}
+host: localhost
+paths:
+  /x:
+    get:
+      responses: {"200": }
+`,
+		"null-response-header": `
+swagger: "2.0"
+info: {title: t, version: "1"}
+host: localhost
+responses:
+  Shared:
+    description: ok
+    headers:
+      X-Shared:
+paths:
+  /x:
+    get:
+      responses:
+        "200":
+          description: ok
+          headers:
+            X-Rate-Limit:
+`,
+	}
+
+	for name, spec := range specs {
+		t.Run(name, func(t *testing.T) {
+			var count int
+			err := New().Parse(strings.NewReader(spec), func(request *types.RequestResponse) bool {
+				count++
+				return false
+			}, "spec.yaml")
+			require.NoError(t, err)
+			if name != "null-path-item" {
+				require.Equal(t, 1, count, "the operation should still produce a request")
+			}
+		})
+	}
+}

pkg/input/formats/testdata/chrome.har

@@ -0,0 +1,71 @@
+{
+  "log": {
+    "version": "1.2",
+    "creator": { "name": "WebInspector", "version": "537.36" },
+    "entries": [
+      {
+        "request": {
+          "method": "GET",
+          "url": "https://ginandjuice.shop/catalog?category=Juice",
+          "httpVersion": "http/2.0",
+          "headers": [
+            { "name": ":authority", "value": "ginandjuice.shop" },
+            { "name": ":method", "value": "GET" },
+            { "name": ":path", "value": "/catalog?category=Juice" },
+            { "name": ":scheme", "value": "https" },
+            { "name": "accept", "value": "text/html" },
+            { "name": "cookie", "value": "session=abc123" }
+          ],
+          "queryString": [{ "name": "category", "value": "Juice" }],
+          "cookies": [],
+          "headersSize": -1,
+          "bodySize": 0
+        },
+        "response": {
+          "status": 200,
+          "statusText": "",
+          "httpVersion": "http/2.0",
+          "headers": [{ "name": "content-type", "value": "text/html; charset=utf-8" }],
+          "content": { "size": 19, "mimeType": "text/html", "text": "PGgxPkp1aWNlPC9oMT4=", "encoding": "base64" }
+        }
+      },
+      {
+        "request": {
+          "method": "POST",
+          "url": "https://ginandjuice.shop/login",
+          "httpVersion": "HTTP/1.1",
+          "headers": [
+            { "name": "Host", "value": "ginandjuice.shop" },
+            { "name": "Content-Type", "value": "application/x-www-form-urlencoded" }
+          ],
+          "postData": { "mimeType": "application/x-www-form-urlencoded", "text": "username=carlos&password=hunter2" }
+        },
+        "response": {
+          "status": 302,
+          "statusText": "Found",
+          "httpVersion": "HTTP/1.1",
+          "headers": [{ "name": "Location", "value": "/my-account" }],
+          "content": { "size": 0, "mimeType": "" }
+        }
+      },
+      {
+        "request": {
+          "method": "GET",
+          "url": "data:image/png;base64,iVBORw0KGgo=",
+          "httpVersion": "",
+          "headers": []
+        },
+        "response": { "status": 200, "statusText": "", "httpVersion": "", "headers": [], "content": { "size": 0 } }
+      },
+      {
+        "request": {
+          "method": "GET",
+          "url": "https://ginandjuice.shop/blocked.js",
+          "httpVersion": "",
+          "headers": []
+        },
+        "response": { "status": 0, "statusText": "", "httpVersion": "", "headers": [], "content": { "size": 0 } }
+      }
+    ]
+  }
+}

pkg/input/formats/testdata/katana.jsonl

@@ -1,6 +0,0 @@
-{"timestamp":"2024-01-01T00:00:00Z","request":{"method":"GET","endpoint":"https://ginandjuice.shop/catalog/product?productId=1","tag":"a","source":"https://ginandjuice.shop/","raw":"GET /catalog/product?productId=1 HTTP/1.1\r\nHost: ginandjuice.shop\r\nUser-Agent: katana\r\n\r\n"}}
-{"timestamp":"2024-01-01T00:00:01Z","request":{"method":"POST","endpoint":"https://ginandjuice.shop/catalog/subscribe","body":"email=test%40test.com","raw":"POST /catalog/subscribe HTTP/1.1\r\nHost: ginandjuice.shop\r\nContent-Type: application/x-www-form-urlencoded\r\nContent-Length: 21\r\n\r\nemail=test%40test.com"}}
-{"timestamp":"2024-01-01T00:00:02Z","request":{"method":"POST","endpoint":"https://ginandjuice.shop/login","headers":{"Content-Type":"application/json"},"body":"{\"username\":\"admin\",\"password\":\"x\"}"}}
-
-not-json-should-be-skipped
-https://ginandjuice.shop/about

pkg/input/provider/http/katana_test.go

@@ -1,35 +0,0 @@
-package http
-
-import (
-	"testing"
-
-	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/contextargs"
-	"github.com/stretchr/testify/require"
-)
-
-func TestHttpInputProviderIteratesAllKatanaRecords(t *testing.T) {
-	input := `{"request":{"method":"GET","endpoint":"https://example.com/a?id=1"}}` + "\n" +
-		`{"request":{"method":"POST","endpoint":"https://example.com/login","body":"user=admin"}}` + "\n" +
-		"https://example.com/bare?q=1\n"
-
-	provider, err := NewHttpInputProvider(&HttpMultiFormatOptions{
-		InputMode:     "katana",
-		InputContents: input,
-	})
-	require.NoError(t, err)
-	require.Equal(t, int64(3), provider.Count())
-
-	var got []string
-	provider.Iterate(func(value *contextargs.MetaInput) bool {
-		got = append(got, value.Input)
-		return true
-	})
-	require.Equal(t, []string{"https://example.com/a?id=1", "https://example.com/login", "https://example.com/bare?q=1"}, got)
-
-	var first []string
-	provider.Iterate(func(value *contextargs.MetaInput) bool {
-		first = append(first, value.Input)
-		return false
-	})
-	require.Len(t, first, 1)
-}

pkg/input/provider/http/multiformat.go

@@ -10,8 +10,8 @@ import (
 	"github.com/projectdiscovery/gologger"
 	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats"
 	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats/burp"
+	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats/har"
 	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats/json"
-	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats/katana"
 	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats/openapi"
 	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats/raw"
 	"github.com/projectdiscovery/nuclei/v3/pkg/input/formats/swagger"
@@ -89,7 +89,7 @@ func NewHttpInputProvider(opts *HttpMultiFormatOptions) (*HttpInputProvider, err
 
 	parseErr := format.Parse(bytes.NewReader(data), func(request *types.RequestResponse) bool {
 		count++
-		return true
+		return false
 	}, opts.InputFile)
 	if parseErr != nil {
 		return nil, errors.Wrap(parseErr, "could not parse input file")
@@ -143,12 +143,12 @@ func (i *HttpInputProvider) Close() {}
 // Supported Providers
 var providersList = []formats.Format{
 	burp.New(),
+	har.New(),
 	json.New(),
 	yaml.New(),
 	openapi.New(),
 	swagger.New(),
 	raw.New(),
-	katana.New(),
 }
 
 // SupportedFormats returns the list of supported formats in comma-separated

pkg/input/provider/list/hmap.go

@@ -89,8 +89,9 @@ func New(opts *Options) (*ListInputProvider, error) {
 		patterns:      normalize.NewFingerprinter(options.MaxURLsPerPattern),
 		ipOptions: &ipOptions{
 			ScanAllIPs: options.ScanAllIPs,
-			IPV4:       sliceutil.Contains(options.IPVersion, "4"),
-			IPV6:       sliceutil.Contains(options.IPVersion, "6"),
+			IPV4:       len(options.IPVersion) == 0 || sliceutil.Contains(options.IPVersion, "4"),
+			IPV6:       len(options.IPVersion) == 0 || sliceutil.Contains(options.IPVersion, "6"),
+			Auto:       len(options.IPVersion) == 0,
 		},
 		excludedHosts: make(map[string]struct{}),
 		profiles:      targetprofile.NewRegistry(configuration.TemplatesDirectory(options, config.DefaultConfig)),
@@ -168,6 +169,61 @@ func (i *ListInputProvider) TargetProfiles() *targetprofile.Registry {
 	return i.profiles
 }
 
+// selectedIPs returns address pins for a hostname. An empty pin delegates
+// DNS and connection fallback to fastdialer, which races IPv4 and IPv6.
+// Set and Del share this selection. With -iv omitted and without -sa, one
+// unpinned target is returned. -sa returns every selected address.
+func (i *ListInputProvider) selectedIPs(executionId, hostname string) []string {
+	if i.ipOptions == nil || (!i.ipOptions.ScanAllIPs && (i.ipOptions.Auto || !i.ipOptions.IPV6)) {
+		return []string{""}
+	}
+	dialers := protocolstate.GetDialersWithId(executionId)
+	if dialers == nil {
+		gologger.Warning().Msgf("cannot resolve %s: dialer %s not found", hostname, executionId)
+		return nil
+	}
+	dnsData, err := dialers.Fastdialer.GetDNSData(hostname)
+	if err != nil {
+		gologger.Warning().Msgf("DNS resolution failed for %s: %v", hostname, err)
+		if i.ipOptions.IPV4 {
+			return []string{""}
+		}
+		return nil
+	}
+	if len(dnsData.A) == 0 && len(dnsData.AAAA) == 0 {
+		gologger.Warning().Msgf("DNS resolution returned no IP addresses for %s", hostname)
+		if i.ipOptions.IPV4 {
+			return []string{""}
+		}
+		return nil
+	}
+	if i.ipOptions.ScanAllIPs {
+		var ips []string
+		if i.ipOptions.IPV6 {
+			ips = append(ips, dnsData.AAAA...)
+		}
+		if i.ipOptions.IPV4 {
+			ips = append(ips, dnsData.A...)
+		}
+		if len(ips) == 0 {
+			gologger.Warning().Msgf("no DNS addresses matching -iv selection for %s", hostname)
+		}
+		return ips
+	}
+	var ips []string
+	if i.ipOptions.IPV6 && len(dnsData.AAAA) > 0 {
+		ips = append(ips, dnsData.AAAA[0])
+	}
+	// Explicit -iv 4,6 scans the first IPv6 address and one unpinned target.
+	if i.ipOptions.IPV4 {
+		ips = append(ips, "")
+	}
+	if len(ips) == 0 {
+		gologger.Warning().Msgf("no IPv6 DNS address found for %s", hostname)
+	}
+	return ips
+}
+
 // Set normalizes and stores passed input values
 func (i *ListInputProvider) Set(executionId string, value string) {
 	i.set(executionId, value, nil)
@@ -202,75 +258,11 @@ func (i *ListInputProvider) set(executionId string, value string, selection *tar
 		return
 	}
 
-	if i.ipOptions.ScanAllIPs {
-		// scan all ips
-		dialers := protocolstate.GetDialersWithId(executionId)
-		if dialers == nil {
-			gologger.Error().Msgf("dialers with executionId %s not found", executionId)
-			return
-		}
-
-		dnsData, err := dialers.Fastdialer.GetDNSData(urlx.Hostname())
-		if err == nil {
-			if (len(dnsData.A) + len(dnsData.AAAA)) > 0 {
-				var ips []string
-				if i.ipOptions.IPV4 {
-					ips = append(ips, dnsData.A...)
-				}
-				if i.ipOptions.IPV6 {
-					ips = append(ips, dnsData.AAAA...)
-				}
-				for _, ip := range ips {
-					if ip == "" {
-						continue
-					}
-					metaInput := contextargs.NewMetaInput()
-					metaInput.Input = URL
-					metaInput.CustomIP = ip
-					i.setItem(metaInput, selection)
-				}
-				return
-			} else {
-				gologger.Debug().Msgf("scanAllIps: no ip's found reverting to default")
-			}
-		} else {
-			// failed to scanallips falling back to defaults
-			gologger.Debug().Msgf("scanAllIps: dns resolution failed: %v", err)
-		}
-	}
-
-	ips := []string{}
-	// only scan the target but ipv6 if it has one
-	if i.ipOptions.IPV6 {
-		dialers := protocolstate.GetDialersWithId(executionId)
-		if dialers == nil {
-			gologger.Error().Msgf("dialers with executionId %s not found", executionId)
-			return
-		}
-
-		dnsData, err := dialers.Fastdialer.GetDNSData(urlx.Hostname())
-		if err == nil && len(dnsData.AAAA) > 0 {
-			// pick/ prefer 1st
-			ips = append(ips, dnsData.AAAA[0])
-		} else {
-			gologger.Warning().Msgf("target does not have ipv6 address falling back to ipv4 %v\n", err)
-		}
-	}
-	if i.ipOptions.IPV4 {
-		// if IPV4 is enabled do not specify ip let dialer handle it
-		ips = append(ips, "")
-	}
-
-	for _, ip := range ips {
+	for _, ip := range i.selectedIPs(executionId, urlx.Hostname()) {
 		metaInput := contextargs.NewMetaInput()
-		if ip != "" {
-			metaInput.Input = URL
-			metaInput.CustomIP = ip
-			i.setItem(metaInput, selection)
-		} else {
-			metaInput.Input = URL
-			i.setItem(metaInput, selection)
-		}
+		metaInput.Input = URL
+		metaInput.CustomIP = ip
+		i.setItem(metaInput, selection)
 	}
 }
 
@@ -493,75 +485,11 @@ func (i *ListInputProvider) Del(executionId string, value string) {
 		return
 	}
 
-	if i.ipOptions.ScanAllIPs {
-		// scan all ips
-		dialers := protocolstate.GetDialersWithId(executionId)
-		if dialers == nil {
-			gologger.Error().Msgf("dialers with executionId %s not found", executionId)
-			return
-		}
-
-		dnsData, err := dialers.Fastdialer.GetDNSData(urlx.Hostname())
-		if err == nil {
-			if (len(dnsData.A) + len(dnsData.AAAA)) > 0 {
-				var ips []string
-				if i.ipOptions.IPV4 {
-					ips = append(ips, dnsData.A...)
-				}
-				if i.ipOptions.IPV6 {
-					ips = append(ips, dnsData.AAAA...)
-				}
-				for _, ip := range ips {
-					if ip == "" {
-						continue
-					}
-					metaInput := contextargs.NewMetaInput()
-					metaInput.Input = value
-					metaInput.CustomIP = ip
-					i.delItem(metaInput)
-				}
-				return
-			} else {
-				gologger.Debug().Msgf("scanAllIps: no ip's found reverting to default")
-			}
-		} else {
-			// failed to scanallips falling back to defaults
-			gologger.Debug().Msgf("scanAllIps: dns resolution failed: %v", err)
-		}
-	}
-
-	ips := []string{}
-	// only scan the target but ipv6 if it has one
-	if i.ipOptions.IPV6 {
-		dialers := protocolstate.GetDialersWithId(executionId)
-		if dialers == nil {
-			gologger.Error().Msgf("dialers with executionId %s not found", executionId)
-			return
-		}
-
-		dnsData, err := dialers.Fastdialer.GetDNSData(urlx.Hostname())
-		if err == nil && len(dnsData.AAAA) > 0 {
-			// pick/ prefer 1st
-			ips = append(ips, dnsData.AAAA[0])
-		} else {
-			gologger.Warning().Msgf("target does not have ipv6 address falling back to ipv4 %v\n", err)
-		}
-	}
-	if i.ipOptions.IPV4 {
-		// if IPV4 is enabled do not specify ip let dialer handle it
-		ips = append(ips, "")
-	}
-
-	for _, ip := range ips {
+	for _, ip := range i.selectedIPs(executionId, urlx.Hostname()) {
 		metaInput := contextargs.NewMetaInput()
-		if ip != "" {
-			metaInput.Input = URL
-			metaInput.CustomIP = ip
-			i.delItem(metaInput)
-		} else {
-			metaInput.Input = URL
-			i.delItem(metaInput)
-		}
+		metaInput.Input = URL
+		metaInput.CustomIP = ip
+		i.delItem(metaInput)
 	}
 }
 

pkg/input/provider/list/ip_version_test.go

@@ -0,0 +1,106 @@
+package list
+
+import (
+	"net"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/miekg/dns"
+	"github.com/projectdiscovery/hmap/store/hybrid"
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/contextargs"
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/protocolstate"
+	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/stretchr/testify/require"
+)
+
+// A local deterministic DNS responder for dual-stack, A-only and AAAA-only hosts.
+type ipVersionDNSHandler struct{}
+
+func (ipVersionDNSHandler) ServeDNS(w dns.ResponseWriter, r *dns.Msg) {
+	reply := new(dns.Msg)
+	reply.SetReply(r)
+	if len(r.Question) == 0 {
+		_ = w.WriteMsg(reply)
+		return
+	}
+	question := r.Question[0]
+	// Some resolvers try search suffixes before the absolute name. Match the
+	// test host prefix rather than answering unrelated names as dual-stack.
+	host := strings.ToLower(question.Name)
+	if strings.HasPrefix(host, "empty.") {
+		_ = w.WriteMsg(reply)
+		return
+	}
+	isDual := strings.HasPrefix(host, "dual.")
+	isV4 := strings.HasPrefix(host, "v4only.")
+	isV6 := strings.HasPrefix(host, "v6only.")
+	if question.Qtype == dns.TypeA && (isDual || isV4) {
+		reply.Answer = append(reply.Answer, &dns.A{Hdr: dns.RR_Header{Name: question.Name, Rrtype: dns.TypeA, Class: dns.ClassINET, Ttl: 60}, A: net.ParseIP("192.0.2.10")})
+	}
+	if question.Qtype == dns.TypeAAAA && (isDual || isV6) {
+		reply.Answer = append(reply.Answer, &dns.AAAA{Hdr: dns.RR_Header{Name: question.Name, Rrtype: dns.TypeAAAA, Class: dns.ClassINET, Ttl: 60}, AAAA: net.ParseIP("2001:db8::10")})
+	}
+	_ = w.WriteMsg(reply)
+}
+
+func TestIPVersionsDefaultAndExplicitSelection(t *testing.T) {
+	conn, err := net.ListenPacket("udp", "127.0.0.1:0")
+	require.NoError(t, err)
+	started := make(chan struct{})
+	server := &dns.Server{PacketConn: conn, Handler: ipVersionDNSHandler{}, NotifyStartedFunc: func() { close(started) }}
+	go func() { _ = server.ActivateAndServe() }()
+	t.Cleanup(func() { _ = server.Shutdown() })
+	select {
+	case <-started:
+	case <-time.After(5 * time.Second):
+		t.Fatal("DNS test resolver did not start")
+	}
+	opts := types.DefaultOptions()
+	opts.ExecutionId = "ip-version-regression"
+	opts.InternalResolversList = []string{conn.LocalAddr().String()}
+	require.NoError(t, protocolstate.Init(opts))
+	t.Cleanup(func() { protocolstate.Close(opts.ExecutionId) })
+	for _, tc := range []struct {
+		name, hostname            string
+		scanAll, ipv4, ipv6, auto bool
+		want                      []string
+	}{
+		{"automatic dual-stack keeps dialer fallback", "dual.test", false, true, true, true, []string{""}},
+		{"automatic IPv6-only keeps dialer fallback", "v6only.test", false, true, true, true, []string{""}},
+		{"auto IPv4 fallback", "v4only.test", false, true, true, true, []string{""}},
+		{"auto empty DNS response fallback", "empty.test", false, true, true, true, []string{""}},
+		{"all empty DNS response fallback", "empty.test", true, true, true, true, []string{""}},
+		{"all dual", "dual.test", true, true, true, true, []string{"192.0.2.10", "2001:db8::10"}},
+		{"all IPv6 only", "v6only.test", true, true, true, true, []string{"2001:db8::10"}},
+		{"all IPv4 only", "v4only.test", true, true, true, true, []string{"192.0.2.10"}},
+		{"explicit IPv4 only", "dual.test", false, true, false, false, []string{""}},
+		{"explicit IPv6 only", "dual.test", false, false, true, false, []string{"2001:db8::10"}},
+		{"explicit both", "dual.test", false, true, true, false, []string{"2001:db8::10", ""}},
+		{"all explicit IPv6", "dual.test", true, false, true, false, []string{"2001:db8::10"}},
+		{"all explicit IPv4", "dual.test", true, true, false, false, []string{"192.0.2.10"}},
+		{"IPv6 restricted on IPv4-only", "v4only.test", true, false, true, false, nil},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			hm, err := hybrid.New(hybrid.DefaultDiskOptions)
+			require.NoError(t, err)
+			input := &ListInputProvider{
+				hostMap:       hm,
+				excludedHosts: make(map[string]struct{}),
+				ipOptions:     &ipOptions{ScanAllIPs: tc.scanAll, IPV4: tc.ipv4, IPV6: tc.ipv6, Auto: tc.auto},
+			}
+			t.Cleanup(input.Close)
+			input.Set(opts.ExecutionId, "https://"+tc.hostname)
+			var got []string
+			input.Iterate(func(value *contextargs.MetaInput) bool {
+				got = append(got, value.CustomIP)
+				return true
+			})
+			require.ElementsMatch(t, tc.want, got)
+			if tc.scanAll && len(tc.want) > 0 {
+				input.Del(opts.ExecutionId, "https://"+tc.hostname)
+				require.EqualValues(t, 0, input.Count(), "Del must remove the same IPs added by Set")
+			}
+		})
+	}
+}

pkg/input/provider/list/utils.go

@@ -4,4 +4,5 @@ type ipOptions struct {
 	ScanAllIPs bool
 	IPV4       bool
 	IPV6       bool
+	Auto       bool // -iv omitted: one unpinned target unless -sa scans every address
 }

pkg/js/compiler/pool.go

@@ -16,10 +16,12 @@ import (
 	syncutil "github.com/projectdiscovery/utils/sync"
 
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libbytes"
+	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libcompress"
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libdcerpc"
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libdcom"
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libfs"
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libgrpc"
+	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libhash"
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libhttp"
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libikev2"
 	_ "github.com/projectdiscovery/nuclei/v3/pkg/js/generated/go/libkerberos"

pkg/js/compiler/stream_libs_test.go

@@ -0,0 +1,158 @@
+package compiler
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"math/rand"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"strconv"
+	"testing"
+	"time"
+
+	"github.com/andybalholm/brotli"
+	"github.com/klauspost/compress/gzip"
+	"github.com/klauspost/compress/zlib"
+	"github.com/klauspost/compress/zstd"
+	"github.com/projectdiscovery/dsl"
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/protocolstate"
+	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	httputil "github.com/projectdiscovery/utils/http"
+	"github.com/stretchr/testify/require"
+)
+
+const streamHashScript = `
+	const net = require('nuclei/net');
+	const compress = require('nuclei/compress');
+	const hash = require('nuclei/hash');
+
+	const conn = net.Open('tcp', %q);
+	conn.Send('GET / HTTP/1.1\r\nHost: test\r\nAccept-Encoding: gzip, deflate, br, zstd\r\n\r\n');
+	let head = '';
+	while (!head.endsWith('\r\n\r\n')) head += conn.RecvString(1);
+	const header = (name) => (head.match(new RegExp('\r\n' + name + ': ([^\r]+)', 'i')) || [])[1];
+	const length = header('Content-Length');
+	const coding = [header('Content-Encoding'), header('Transfer-Encoding')].filter(Boolean).join(', ');
+
+	const d = compress.NewDecoder(coding);
+	const sha = hash.New('sha256'), mmh3 = hash.New('mmh3');
+	const take = (out) => { sha.Write(out); mmh3.Write(out); };
+	if (length !== undefined) {
+		for (let read = 0; read < parseInt(length); ) {
+			const chunk = conn.Recv(65536);
+			read += chunk.length;
+			for (let out = d.Write(chunk); out.length > 0; out = d.Read()) take(out);
+		}
+		for (let out = d.Close(); out.length > 0; out = d.Read()) take(out);
+	} else {
+		while (!d.Finished()) {
+			for (let out = d.Write(conn.Recv(65536)); out.length > 0; out = d.Read()) take(out);
+		}
+	}
+	ExportAs('sha256', sha.Sum());
+	ExportAs('mmh3', mmh3.Sum());
+	true;
+`
+
+func encodeBody(t *testing.T, encoding string, data []byte) []byte {
+	t.Helper()
+	var buf bytes.Buffer
+	var w io.WriteCloser
+	switch encoding {
+	case "gzip":
+		w = gzip.NewWriter(&buf)
+	case "deflate":
+		w = zlib.NewWriter(&buf)
+	case "br":
+		w = brotli.NewWriter(&buf)
+	case "zstd":
+		zw, err := zstd.NewWriter(&buf)
+		require.NoError(t, err)
+		w = zw
+	case "":
+		return data
+	}
+	_, err := w.Write(data)
+	require.NoError(t, err)
+	require.NoError(t, w.Close())
+	return buf.Bytes()
+}
+
+func TestStreamingHashMatchesHTTPEngine(t *testing.T) {
+	rng := rand.New(rand.NewSource(1))
+	var body []byte
+	for len(body) < 6<<20 {
+		body = append(body, "large compressed response body "...)
+		body = append(body, byte(rng.Intn(256)))
+	}
+
+	executionID := t.Name()
+	require.NoError(t, protocolstate.Init(&types.Options{ExecutionId: executionID}))
+	t.Cleanup(func() { protocolstate.Close(executionID) })
+
+	for _, tc := range []struct {
+		encoding string
+		chunked  bool
+	}{
+		{"gzip", false},
+		{"gzip", true},
+		{"deflate", false},
+		{"br", true},
+		{"zstd", true},
+		{"", true},
+	} {
+		t.Run(fmt.Sprintf("%s chunked=%v", tc.encoding, tc.chunked), func(t *testing.T) {
+			encoded := encodeBody(t, tc.encoding, body)
+			server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+				if tc.encoding != "" {
+					w.Header().Set("Content-Encoding", tc.encoding)
+				}
+				if !tc.chunked {
+					w.Header().Set("Content-Length", strconv.Itoa(len(encoded)))
+					_, _ = w.Write(encoded)
+					return
+				}
+				// Flushing before the handler returns makes the body chunked.
+				for rest := encoded; len(rest) > 0; {
+					n := min(len(rest), 50000)
+					_, _ = w.Write(rest[:n])
+					w.(http.Flusher).Flush()
+					rest = rest[n:]
+				}
+			}))
+			defer server.Close()
+
+			client := &http.Client{Transport: &http.Transport{DisableCompression: true}}
+			resp, err := client.Get(server.URL)
+			require.NoError(t, err)
+			if tc.chunked {
+				require.Equal(t, []string{"chunked"}, resp.TransferEncoding)
+			}
+			chain := httputil.NewResponseChain(resp, 64<<20)
+			defer chain.Close()
+			require.NoError(t, chain.Fill())
+			require.Equal(t, len(body), chain.Body().Len(), "the engine must decode the whole body")
+			engineBody := map[string]interface{}{"body": chain.BodyString()}
+			wantSha, err := dsl.EvalExpr("sha256(body)", engineBody)
+			require.NoError(t, err)
+			wantMmh3, err := dsl.EvalExpr("mmh3(body)", engineBody)
+			require.NoError(t, err)
+
+			serverURL, err := url.Parse(server.URL)
+			require.NoError(t, err)
+			script := fmt.Sprintf(streamHashScript, serverURL.Host)
+			compiled, err := SourceAutoMode(script, false)
+			require.NoError(t, err)
+			result, err := New().ExecuteWithOptions(t.Context(), compiled, NewExecuteArgs(), &ExecuteOptions{
+				ExecutionId:     executionID,
+				Source:          &script,
+				TimeoutVariants: &types.Timeouts{JsCompilerExecutionTimeout: 20 * time.Second},
+			})
+			require.NoError(t, err)
+			require.Equal(t, fmt.Sprint(wantSha), result["sha256"])
+			require.Equal(t, fmt.Sprint(wantMmh3), result["mmh3"])
+		})
+	}
+}

pkg/js/generated/go/libcompress/compress.go

@@ -0,0 +1,30 @@
+package compress
+
+import (
+	lib_compress "github.com/projectdiscovery/nuclei/v3/pkg/js/libs/compress"
+
+	"github.com/projectdiscovery/goja"
+	"github.com/projectdiscovery/nuclei/v3/pkg/js/gojs"
+)
+
+var (
+	module = gojs.NewGojaModule("nuclei/compress")
+)
+
+func init() {
+	module.Set(
+		gojs.Objects{
+			// Functions
+			"NewDecoder": lib_compress.NewDecoder,
+
+			// Var and consts
+
+			// Objects / Classes
+			"Decoder": gojs.GetClassConstructor[lib_compress.Decoder](&lib_compress.Decoder{}),
+		},
+	).Register()
+}
+
+func Enable(runtime *goja.Runtime) {
+	module.Enable(runtime)
+}

pkg/js/generated/go/libhash/hash.go

@@ -0,0 +1,30 @@
+package hash
+
+import (
+	lib_hash "github.com/projectdiscovery/nuclei/v3/pkg/js/libs/hash"
+
+	"github.com/projectdiscovery/goja"
+	"github.com/projectdiscovery/nuclei/v3/pkg/js/gojs"
+)
+
+var (
+	module = gojs.NewGojaModule("nuclei/hash")
+)
+
+func init() {
+	module.Set(
+		gojs.Objects{
+			// Functions
+			"New": lib_hash.New,
+
+			// Var and consts
+
+			// Objects / Classes
+			"Hash": gojs.GetClassConstructor[lib_hash.Hash](&lib_hash.Hash{}),
+		},
+	).Register()
+}
+
+func Enable(runtime *goja.Runtime) {
+	module.Enable(runtime)
+}

pkg/js/generated/ts/compress.ts

@@ -0,0 +1,102 @@
+
+
+/**
+ * NewDecoder returns a decoder for gzip, deflate, br, zstd, chunked or
+ * identity. List several comma-separated, in the order they were applied.
+ * @example
+ * ```javascript
+ * const compress = require('nuclei/compress');
+ * const d = compress.NewDecoder('br');
+ * // Content-Encoding: gzip sent with Transfer-Encoding: chunked
+ * const c = compress.NewDecoder('gzip, chunked');
+ * ```
+ */
+export function NewDecoder(encoding: string): Decoder | null {
+    return null;
+}
+
+
+
+/**
+ * Decoder decodes a compressed stream written to it in chunks. Each call
+ * returns up to about 1 MB; call Read until it returns nothing before the
+ * next Write.
+ * @example
+ * ```javascript
+ * const compress = require('nuclei/compress');
+ * const hash = require('nuclei/hash');
+ * const d = compress.NewDecoder('gzip');
+ * const h = hash.New('sha256');
+ * for (const chunk of chunks) {
+ *   for (let out = d.Write(chunk); out.length > 0; out = d.Read()) h.Write(out);
+ * }
+ * for (let out = d.Close(); out.length > 0; out = d.Read()) h.Write(out);
+ * log(h.Sum());
+ * ```
+ */
+export class Decoder {
+    
+
+    // Constructor of Decoder
+    constructor() {}
+    /**
+    * Write decodes the next chunk and returns up to about 1 MB of output.
+    * @example
+    * ```javascript
+    * const compress = require('nuclei/compress');
+    * const d = compress.NewDecoder('gzip');
+    * const out = d.Write(chunk);
+    * ```
+    */
+    public Write(chunk: Uint8Array): Uint8Array | null {
+        return null;
+    }
+    
+
+    /**
+    * Read returns more decoded bytes, or none once the decoder needs input.
+    * @example
+    * ```javascript
+    * const compress = require('nuclei/compress');
+    * const d = compress.NewDecoder('gzip');
+    * for (let out = d.Write(chunk); out.length > 0; out = d.Read()) log(out.length);
+    * ```
+    */
+    public Read(): Uint8Array | null {
+        return null;
+    }
+    
+
+    /**
+    * Close ends the input and returns the last decoded bytes. A truncated
+    * stream errors once the bytes before the cut have been returned.
+    * @example
+    * ```javascript
+    * const compress = require('nuclei/compress');
+    * const d = compress.NewDecoder('gzip');
+    * d.Write(chunk);
+    * for (let out = d.Close(); out.length > 0; out = d.Read()) log(out.length);
+    * ```
+    */
+    public Close(): Uint8Array | null {
+        return null;
+    }
+    
+
+    /**
+    * Finished reports whether a stream that marks its own end, such as a
+    * chunked body, has ended and all its output was read.
+    * @example
+    * ```javascript
+    * const compress = require('nuclei/compress');
+    * const d = compress.NewDecoder('gzip, chunked');
+    * while (!d.Finished()) for (let out = d.Write(conn.Recv(65536)); out.length > 0; out = d.Read()) log(out.length);
+    * ```
+    */
+    public Finished(): boolean {
+        return false;
+    }
+    
+
+}
+

pkg/js/generated/ts/hash.ts

@@ -0,0 +1,77 @@
+
+
+/**
+ * New returns a running hash for md5, sha1, sha256, sha512 or mmh3. Sum
+ * matches the DSL function of the same name.
+ * @example
+ * ```javascript
+ * const hash = require('nuclei/hash');
+ * const h = hash.New('sha256');
+ * ```
+ */
+export function New(algorithm: string): Hash | null {
+    return null;
+}
+
+
+
+/**
+ * Hash is a running hash.
+ * @example
+ * ```javascript
+ * const hash = require('nuclei/hash');
+ * const h = hash.New('sha256');
+ * h.WriteString('hello');
+ * log(h.Sum());
+ * ```
+ */
+export class Hash {
+    
+
+    // Constructor of Hash
+    constructor() {}
+    /**
+    * Write adds data to the hash.
+    * @example
+    * ```javascript
+    * const hash = require('nuclei/hash');
+    * const h = hash.New('sha256');
+    * h.Write([104, 105]);
+    * ```
+    */
+    public Write(data: Uint8Array): void {
+        return;
+    }
+    
+
+    /**
+    * WriteString adds the string data to the hash.
+    * @example
+    * ```javascript
+    * const hash = require('nuclei/hash');
+    * const h = hash.New('sha256');
+    * h.WriteString('hi');
+    * ```
+    */
+    public WriteString(data: string): void {
+        return;
+    }
+    
+
+    /**
+    * Sum returns the hash of the data written so far, formatted as the DSL does.
+    * @example
+    * ```javascript
+    * const hash = require('nuclei/hash');
+    * const h = hash.New('mmh3');
+    * h.WriteString('hello');
+    * log(h.Sum());
+    * ```
+    */
+    public Sum(): string | null {
+        return null;
+    }
+    
+
+}
+

pkg/js/generated/ts/index.ts

@@ -1,9 +1,11 @@
 export * as bytes from './bytes';
+export * as compress from './compress';
 export * as dcom from './dcom';
 export * as dcerpc from './dcerpc';
 export * as fs from './fs';
 export * as goconsole from './goconsole';
 export * as grpc from './grpc';
+export * as hash from './hash';
 export * as http from './http';
 export * as ikev2 from './ikev2';
 export * as kerberos from './kerberos';

pkg/js/libs/compress/decoder.go

@@ -0,0 +1,303 @@
+// Package compress exposes streaming decompression to nuclei JavaScript templates.
+package compress
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"net/http/httputil"
+	"slices"
+	"strings"
+
+	"github.com/dsnet/compress/brotli"
+	"github.com/klauspost/compress/gzip"
+	"github.com/klauspost/compress/zlib"
+	"github.com/klauspost/compress/zstd"
+)
+
+// maxDecodedChunk caps the decoded bytes one call hands back.
+const maxDecodedChunk = 1 << 20
+
+type pauseReason int
+
+const (
+	needInput pauseReason = iota
+	outputFull
+	finished
+)
+
+type (
+	// Decoder decodes a compressed stream written to it in chunks. Each call
+	// returns up to about 1 MB; call Read until it returns nothing before the
+	// next Write.
+	// @example
+	// ```javascript
+	// const compress = require('nuclei/compress');
+	// const hash = require('nuclei/hash');
+	// const d = compress.NewDecoder('gzip');
+	// const h = hash.New('sha256');
+	// for (const chunk of chunks) {
+	//   for (let out = d.Write(chunk); out.length > 0; out = d.Read()) h.Write(out);
+	// }
+	// for (let out = d.Close(); out.length > 0; out = d.Read()) h.Write(out);
+	// log(h.Sum());
+	// ```
+	Decoder struct {
+		ctx context.Context
+		// The decode goroutine and the template's calls take turns over these
+		// channels, so no field is accessed by both at once.
+		resume chan []byte
+		paused chan pauseReason
+		done   chan struct{}
+		out    []byte
+
+		state      pauseReason
+		inputEnded bool
+
+		pending []byte
+		eof     bool
+		err     error
+	}
+)
+
+// NewDecoder returns a decoder for gzip, deflate, br, zstd, chunked or
+// identity. List several comma-separated, in the order they were applied.
+// @example
+// ```javascript
+// const compress = require('nuclei/compress');
+// const d = compress.NewDecoder('br');
+// // Content-Encoding: gzip sent with Transfer-Encoding: chunked
+// const c = compress.NewDecoder('gzip, chunked');
+// ```
+func NewDecoder(ctx context.Context, encoding string) (*Decoder, error) {
+	var opens []func(io.Reader) (io.Reader, func(), error)
+	for name := range strings.SplitSeq(encoding, ",") {
+		switch name = strings.ToLower(strings.TrimSpace(name)); name {
+		case "gzip":
+			opens = append(opens, func(r io.Reader) (io.Reader, func(), error) {
+				zr, err := gzip.NewReader(r)
+				return zr, func() {}, err
+			})
+		case "deflate":
+			opens = append(opens, func(r io.Reader) (io.Reader, func(), error) {
+				zr, err := zlib.NewReader(r)
+				return zr, func() {}, err
+			})
+		case "br":
+			opens = append(opens, func(r io.Reader) (io.Reader, func(), error) {
+				br, err := brotli.NewReader(r, nil)
+				return br, func() {}, err
+			})
+		case "zstd":
+			opens = append(opens, func(r io.Reader) (io.Reader, func(), error) {
+				// One goroutine so no background workers outlive the stream.
+				zr, err := zstd.NewReader(r, zstd.WithDecoderConcurrency(1))
+				if err != nil {
+					return nil, nil, err
+				}
+				return zr, zr.Close, nil
+			})
+		case "chunked":
+			opens = append(opens, func(r io.Reader) (io.Reader, func(), error) {
+				return httputil.NewChunkedReader(r), func() {}, nil
+			})
+		case "", "identity":
+		default:
+			return nil, fmt.Errorf("unsupported encoding %q, expected gzip, deflate, br, zstd, chunked or identity", name)
+		}
+	}
+	slices.Reverse(opens)
+	open := func(r io.Reader) (io.Reader, func(), error) {
+		var closers []func()
+		closeAll := func() {
+			for _, c := range slices.Backward(closers) {
+				c()
+			}
+		}
+		for _, o := range opens {
+			next, c, err := o(r)
+			if err != nil {
+				closeAll()
+				return nil, nil, err
+			}
+			r = next
+			closers = append(closers, c)
+		}
+		return r, closeAll, nil
+	}
+	d := &Decoder{
+		ctx:    ctx,
+		resume: make(chan []byte),
+		paused: make(chan pauseReason),
+		done:   make(chan struct{}),
+	}
+	go d.run(open)
+	d.wait()
+	return d, nil
+}
+
+// Write decodes the next chunk and returns up to about 1 MB of output.
+// @example
+// ```javascript
+// const compress = require('nuclei/compress');
+// const d = compress.NewDecoder('gzip');
+// const out = d.Write(chunk);
+// ```
+func (d *Decoder) Write(chunk []byte) ([]byte, error) {
+	if d.done == nil {
+		return nil, errNotCreatedByNewDecoder
+	}
+	if d.inputEnded {
+		return nil, errors.New("decoder is closed")
+	}
+	if d.state == outputFull {
+		return nil, errors.New("decoded bytes are waiting: call Read until it returns no bytes before writing more")
+	}
+	if d.state == needInput && len(chunk) > 0 {
+		// The decode goroutine may still hold the chunk after Write returns.
+		d.resumeWith(bytes.Clone(chunk))
+	}
+	return d.result()
+}
+
+// Read returns more decoded bytes, or none once the decoder needs input.
+// @example
+// ```javascript
+// const compress = require('nuclei/compress');
+// const d = compress.NewDecoder('gzip');
+// for (let out = d.Write(chunk); out.length > 0; out = d.Read()) log(out.length);
+// ```
+func (d *Decoder) Read() ([]byte, error) {
+	if d.done == nil {
+		return nil, errNotCreatedByNewDecoder
+	}
+	d.advance()
+	return d.result()
+}
+
+// Close ends the input and returns the last decoded bytes. A truncated
+// stream errors once the bytes before the cut have been returned.
+// @example
+// ```javascript
+// const compress = require('nuclei/compress');
+// const d = compress.NewDecoder('gzip');
+// d.Write(chunk);
+// for (let out = d.Close(); out.length > 0; out = d.Read()) log(out.length);
+// ```
+func (d *Decoder) Close() ([]byte, error) {
+	if d.done == nil {
+		return nil, errNotCreatedByNewDecoder
+	}
+	d.inputEnded = true
+	d.advance()
+	return d.result()
+}
+
+// Finished reports whether a stream that marks its own end, such as a
+// chunked body, has ended and all its output was read.
+// @example
+// ```javascript
+// const compress = require('nuclei/compress');
+// const d = compress.NewDecoder('gzip, chunked');
+// while (!d.Finished()) for (let out = d.Write(conn.Recv(65536)); out.length > 0; out = d.Read()) log(out.length);
+// ```
+func (d *Decoder) Finished() bool {
+	return d.state == finished && len(d.out) == 0
+}
+
+var errNotCreatedByNewDecoder = errors.New("decoder was not created with compress.NewDecoder")
+
+func (d *Decoder) advance() {
+	for len(d.out) == 0 && (d.state == outputFull || d.state == needInput && d.inputEnded) {
+		d.resumeWith(nil)
+	}
+}
+
+func (d *Decoder) result() ([]byte, error) {
+	out := d.out
+	d.out = nil
+	if len(out) == 0 && d.state == finished {
+		return nil, d.err
+	}
+	return out, nil
+}
+
+func (d *Decoder) resumeWith(chunk []byte) {
+	select {
+	case d.resume <- chunk:
+		d.wait()
+	case <-d.done:
+		d.state = finished
+	}
+}
+
+func (d *Decoder) wait() {
+	select {
+	case d.state = <-d.paused:
+	case <-d.done:
+		d.state = finished
+	}
+}
+
+func (d *Decoder) run(open func(io.Reader) (io.Reader, func(), error)) {
+	defer close(d.done)
+	r, closeReader, err := open(decoderInput{d})
+	if err != nil {
+		d.err = err
+		return
+	}
+	defer closeReader()
+	_, err = io.Copy(decoderOutput{d}, r)
+	d.err = err
+}
+
+func (d *Decoder) pause(reason pauseReason) ([]byte, error) {
+	select {
+	case d.paused <- reason:
+	case <-d.ctx.Done():
+		return nil, context.Cause(d.ctx)
+	}
+	select {
+	case chunk := <-d.resume:
+		return chunk, nil
+	case <-d.ctx.Done():
+		return nil, context.Cause(d.ctx)
+	}
+}
+
+type decoderInput struct{ d *Decoder }
+
+func (in decoderInput) Read(p []byte) (int, error) {
+	d := in.d
+	for len(d.pending) == 0 {
+		if d.eof {
+			return 0, io.EOF
+		}
+		chunk, err := d.pause(needInput)
+		if err != nil {
+			return 0, err
+		}
+		if chunk == nil {
+			d.eof = true
+		}
+		d.pending = chunk
+	}
+	n := copy(p, d.pending)
+	d.pending = d.pending[n:]
+	return n, nil
+}
+
+type decoderOutput struct{ d *Decoder }
+
+func (out decoderOutput) Write(p []byte) (int, error) {
+	d := out.d
+	d.out = append(d.out, p...)
+	if len(d.out) >= maxDecodedChunk {
+		if _, err := d.pause(outputFull); err != nil {
+			return 0, err
+		}
+	}
+	return len(p), nil
+}

pkg/js/libs/compress/decoder_test.go

@@ -0,0 +1,192 @@
+package compress
+
+import (
+	"bytes"
+	"context"
+	"io"
+	"math/rand"
+	"net/http/httputil"
+	"testing"
+	"time"
+
+	"github.com/andybalholm/brotli"
+	"github.com/klauspost/compress/gzip"
+	"github.com/klauspost/compress/zlib"
+	"github.com/klauspost/compress/zstd"
+	"github.com/stretchr/testify/require"
+)
+
+func compressed(t *testing.T, encoding string, data []byte) []byte {
+	t.Helper()
+	var buf bytes.Buffer
+	var w io.WriteCloser
+	switch encoding {
+	case "gzip":
+		w = gzip.NewWriter(&buf)
+	case "deflate":
+		w = zlib.NewWriter(&buf)
+	case "br":
+		w = brotli.NewWriter(&buf)
+	case "zstd":
+		zw, err := zstd.NewWriter(&buf)
+		require.NoError(t, err)
+		w = zw
+	case "identity":
+		return data
+	}
+	_, err := w.Write(data)
+	require.NoError(t, err)
+	require.NoError(t, w.Close())
+	return buf.Bytes()
+}
+
+func decodeInChunks(t *testing.T, d *Decoder, input []byte) ([]byte, error) {
+	t.Helper()
+	rng := rand.New(rand.NewSource(1))
+	var decoded []byte
+	collect := func(out []byte, err error) bool {
+		require.LessOrEqual(t, len(out), maxDecodedChunk+64<<10)
+		decoded = append(decoded, out...)
+		return err == nil && len(out) > 0
+	}
+	drain := func(out []byte, err error) error {
+		for collect(out, err) {
+			out, err = d.Read()
+		}
+		return err
+	}
+	for len(input) > 0 {
+		n := min(len(input), 1+rng.Intn(70000))
+		out, err := d.Write(input[:n])
+		if err := drain(out, err); err != nil {
+			return decoded, err
+		}
+		input = input[n:]
+	}
+	out, err := d.Close()
+	return decoded, drain(out, err)
+}
+
+func TestDecoderMatchesWholeBodyDecode(t *testing.T) {
+	data := bytes.Repeat([]byte("nuclei streaming decoder "), 400000)
+	for _, encoding := range []string{"gzip", "deflate", "br", "zstd", "identity"} {
+		t.Run(encoding, func(t *testing.T) {
+			d, err := NewDecoder(t.Context(), encoding)
+			require.NoError(t, err)
+			decoded, err := decodeInChunks(t, d, compressed(t, encoding, data))
+			require.NoError(t, err)
+			require.Equal(t, data, decoded)
+		})
+	}
+}
+
+func chunked(t *testing.T, data []byte) []byte {
+	t.Helper()
+	var buf bytes.Buffer
+	w := httputil.NewChunkedWriter(&buf)
+	for len(data) > 0 {
+		n := min(len(data), 8000)
+		_, err := w.Write(data[:n])
+		require.NoError(t, err)
+		data = data[n:]
+	}
+	require.NoError(t, w.Close())
+	buf.WriteString("\r\n")
+	return buf.Bytes()
+}
+
+func TestDecoderDecodesChunkedTransfer(t *testing.T) {
+	data := bytes.Repeat([]byte("nuclei chunked decoder "), 400000)
+	for _, encoding := range []string{"gzip", "deflate", "br", "zstd", "identity"} {
+		t.Run(encoding, func(t *testing.T) {
+			d, err := NewDecoder(t.Context(), encoding+", chunked")
+			require.NoError(t, err)
+			decoded, err := decodeInChunks(t, d, chunked(t, compressed(t, encoding, data)))
+			require.NoError(t, err)
+			require.Equal(t, data, decoded)
+		})
+	}
+}
+
+func TestDecoderFinishesAtLastChunk(t *testing.T) {
+	data := bytes.Repeat([]byte("finished "), 1000)
+	d, err := NewDecoder(t.Context(), "gzip, chunked")
+	require.NoError(t, err)
+	input := chunked(t, compressed(t, "gzip", data))
+
+	half := len(input) / 2
+	_, err = d.Write(input[:half])
+	require.NoError(t, err)
+	require.False(t, d.Finished())
+
+	out, err := d.Write(input[half:])
+	require.NoError(t, err)
+	var decoded []byte
+	for ; len(out) > 0; out, err = d.Read() {
+		decoded = append(decoded, out...)
+	}
+	require.NoError(t, err)
+	require.Equal(t, data, decoded)
+	require.True(t, d.Finished(), "decoder must finish without Close once the last chunk arrives")
+}
+
+func TestDecoderAppliesCodingsInReverse(t *testing.T) {
+	data := bytes.Repeat([]byte("two codings "), 100000)
+	input := compressed(t, "br", compressed(t, "gzip", data))
+	d, err := NewDecoder(t.Context(), "gzip, br")
+	require.NoError(t, err)
+	decoded, err := decodeInChunks(t, d, input)
+	require.NoError(t, err)
+	require.Equal(t, data, decoded)
+}
+
+func TestDecoderReportsTruncatedStream(t *testing.T) {
+	data := bytes.Repeat([]byte("truncated "), 200000)
+	for _, encoding := range []string{"gzip", "deflate", "br", "zstd", "identity, chunked"} {
+		t.Run(encoding, func(t *testing.T) {
+			var input []byte
+			if encoding == "identity, chunked" {
+				input = chunked(t, data)
+			} else {
+				input = compressed(t, encoding, data)
+			}
+			d, err := NewDecoder(t.Context(), encoding)
+			require.NoError(t, err)
+			decoded, err := decodeInChunks(t, d, input[:len(input)/2])
+			require.Error(t, err)
+			require.True(t, bytes.HasPrefix(data, decoded), "bytes before the cut must still be handed back")
+		})
+	}
+}
+
+func TestDecoderStopsWhenContextEnds(t *testing.T) {
+	ctx, cancel := context.WithCancel(t.Context())
+	d, err := NewDecoder(ctx, "gzip")
+	require.NoError(t, err)
+	_, err = d.Write(compressed(t, "gzip", []byte("abandoned"))[:5])
+	require.NoError(t, err)
+
+	cancel()
+	select {
+	case <-d.done:
+	case <-time.After(time.Second):
+		t.Fatal("decode goroutine outlived its context")
+	}
+}
+
+func TestNewDecoderRejectsUnknownEncoding(t *testing.T) {
+	_, err := NewDecoder(t.Context(), "compress")
+	require.Error(t, err)
+	_, err = NewDecoder(t.Context(), "gzip, compress")
+	require.Error(t, err)
+}
+
+func TestDecoderNotCreatedByNewDecoderErrors(t *testing.T) {
+	var d Decoder
+	_, err := d.Write([]byte("x"))
+	require.ErrorIs(t, err, errNotCreatedByNewDecoder)
+	_, err = d.Read()
+	require.ErrorIs(t, err, errNotCreatedByNewDecoder)
+	_, err = d.Close()
+	require.ErrorIs(t, err, errNotCreatedByNewDecoder)
+}

pkg/js/libs/hash/hash.go

@@ -0,0 +1,100 @@
+// Package hash exposes incremental hashes to nuclei JavaScript templates.
+package hash
+
+import (
+	"crypto/md5"
+	"crypto/sha1"
+	"crypto/sha256"
+	"crypto/sha512"
+	"encoding/hex"
+	"errors"
+	"fmt"
+	"hash"
+	"strings"
+
+	"github.com/spaolacci/murmur3"
+)
+
+type (
+	// Hash is a running hash.
+	// @example
+	// ```javascript
+	// const hash = require('nuclei/hash');
+	// const h = hash.New('sha256');
+	// h.WriteString('hello');
+	// log(h.Sum());
+	// ```
+	Hash struct {
+		h    hash.Hash
+		mmh3 bool
+	}
+)
+
+// New returns a running hash for md5, sha1, sha256, sha512 or mmh3. Sum
+// matches the DSL function of the same name.
+// @example
+// ```javascript
+// const hash = require('nuclei/hash');
+// const h = hash.New('sha256');
+// ```
+func New(algorithm string) (*Hash, error) {
+	switch strings.ToLower(algorithm) {
+	case "md5":
+		return &Hash{h: md5.New()}, nil
+	case "sha1":
+		return &Hash{h: sha1.New()}, nil
+	case "sha256":
+		return &Hash{h: sha256.New()}, nil
+	case "sha512":
+		return &Hash{h: sha512.New()}, nil
+	case "mmh3":
+		return &Hash{h: murmur3.New32WithSeed(0), mmh3: true}, nil
+	}
+	return nil, fmt.Errorf("unsupported hash algorithm %q, expected md5, sha1, sha256, sha512 or mmh3", algorithm)
+}
+
+// Write adds data to the hash.
+// @example
+// ```javascript
+// const hash = require('nuclei/hash');
+// const h = hash.New('sha256');
+// h.Write([104, 105]);
+// ```
+func (h *Hash) Write(data []byte) error {
+	if h.h == nil {
+		return errNotCreatedByNew
+	}
+	_, _ = h.h.Write(data)
+	return nil
+}
+
+// WriteString adds the string data to the hash.
+// @example
+// ```javascript
+// const hash = require('nuclei/hash');
+// const h = hash.New('sha256');
+// h.WriteString('hi');
+// ```
+func (h *Hash) WriteString(data string) error {
+	return h.Write([]byte(data))
+}
+
+// Sum returns the hash of the data written so far, formatted as the DSL does.
+// @example
+// ```javascript
+// const hash = require('nuclei/hash');
+// const h = hash.New('mmh3');
+// h.WriteString('hello');
+// log(h.Sum());
+// ```
+func (h *Hash) Sum() (string, error) {
+	if h.h == nil {
+		return "", errNotCreatedByNew
+	}
+	if h.mmh3 {
+		return fmt.Sprintf("%d", int32(h.h.(hash.Hash32).Sum32())), nil
+	}
+	return hex.EncodeToString(h.h.Sum(nil)), nil
+}
+
+var errNotCreatedByNew = errors.New("hash was not created with hash.New")

pkg/js/libs/hash/hash_test.go

@@ -0,0 +1,43 @@
+package hash
+
+import (
+	"testing"
+
+	"github.com/projectdiscovery/dsl"
+	"github.com/stretchr/testify/require"
+)
+
+func TestSumMatchesDSLAcrossChunks(t *testing.T) {
+	data := make([]byte, 3<<20+7)
+	for i := range data {
+		data[i] = byte(i * 31)
+	}
+	for _, algorithm := range []string{"md5", "sha1", "sha256", "sha512", "mmh3"} {
+		t.Run(algorithm, func(t *testing.T) {
+			h, err := New(algorithm)
+			require.NoError(t, err)
+			for chunk := data; len(chunk) > 0; {
+				n := min(len(chunk), 65521)
+				require.NoError(t, h.Write(chunk[:n]))
+				chunk = chunk[n:]
+			}
+			want, err := dsl.EvalExpr(algorithm+"(body)", map[string]interface{}{"body": string(data)})
+			require.NoError(t, err)
+			sum, err := h.Sum()
+			require.NoError(t, err)
+			require.Equal(t, want, sum)
+		})
+	}
+}
+
+func TestNewRejectsUnknownAlgorithm(t *testing.T) {
+	_, err := New("crc32")
+	require.Error(t, err)
+}
+
+func TestHashNotCreatedByNewErrors(t *testing.T) {
+	var h Hash
+	require.ErrorIs(t, h.WriteString("x"), errNotCreatedByNew)
+	_, err := h.Sum()
+	require.ErrorIs(t, err, errNotCreatedByNew)
+}

pkg/js/libs/mysql/mysql_fingerprint_client_test.go

@@ -2,7 +2,11 @@ package mysql
 
 import (
 	"context"
+	"fmt"
 	"net"
+	"os"
+	"path/filepath"
+	"runtime"
 	"strings"
 	"sync/atomic"
 	"testing"
@@ -169,6 +173,84 @@ func TestFingerprintMySQLDeniedByNetworkPolicy(t *testing.T) {
 	require.Contains(t, err.Error(), "127.0.0.1")
 }
 
+func TestConnectWithDSNDeniedByNetworkPolicy(t *testing.T) {
+	t.Parallel()
+
+	executionID := "mysql-dsn-deny-" + strings.NewReplacer("/", "-", " ", "-").Replace(t.Name())
+	require.NoError(t, protocolstate.Init(&types.Options{
+		ExecutionId:                executionID,
+		RestrictLocalNetworkAccess: true,
+	}))
+	t.Cleanup(func() { protocolstate.Close(executionID) })
+	ctx := context.WithValue(context.Background(), "executionId", executionID) //nolint:staticcheck
+
+	var accepted atomic.Int32
+	ln, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = ln.Close() })
+	go func() {
+		for {
+			conn, err := ln.Accept()
+			if err != nil {
+				return
+			}
+			accepted.Add(1)
+			_ = conn.Close()
+		}
+	}()
+
+	port := ln.Addr().(*net.TCPAddr).Port
+	dsn := fmt.Sprintf("vuser:vpass@tcp(127.0.0.1:%d)/vdb", port)
+	ok, err := (&MySQLClient{}).ConnectWithDSN(ctx, dsn)
+	require.Error(t, err)
+	require.False(t, ok)
+	require.Contains(t, err.Error(), "127.0.0.1")
+	require.Zero(t, accepted.Load())
+}
+
+func TestConnectWithDSNUnixDeniedByNetworkPolicy(t *testing.T) {
+	t.Parallel()
+
+	executionID := "mysql-dsn-unix-deny-" + strings.NewReplacer("/", "-", " ", "-").Replace(t.Name())
+	require.NoError(t, protocolstate.Init(&types.Options{
+		ExecutionId:                executionID,
+		RestrictLocalNetworkAccess: true,
+	}))
+	t.Cleanup(func() { protocolstate.Close(executionID) })
+	ctx := context.WithValue(context.Background(), "executionId", executionID) //nolint:staticcheck
+
+	// t.TempDir can exceed the unix socket path limit on macOS, so use a short dir.
+	dir, err := os.MkdirTemp("", "nuclei-mysql")
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = os.RemoveAll(dir) })
+	sock := filepath.Join(dir, "lna.sock")
+	ln, err := net.Listen("unix", sock)
+	if err != nil && runtime.GOOS == "windows" {
+		t.Skipf("unix sockets not available: %v", err)
+	}
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = ln.Close() })
+
+	var accepted atomic.Int32
+	go func() {
+		for {
+			conn, err := ln.Accept()
+			if err != nil {
+				return
+			}
+			accepted.Add(1)
+			_ = conn.Close()
+		}
+	}()
+
+	dsn := fmt.Sprintf("vuser:vpass@unix(%s)/vdb", sock)
+	ok, err := (&MySQLClient{}).ConnectWithDSN(ctx, dsn)
+	require.Error(t, err)
+	require.False(t, ok)
+	require.Contains(t, err.Error(), "127.0.0.1")
+	require.Zero(t, accepted.Load())
+}
+
 func TestFingerprintMySQLVersionNeverEmptyOnHandshakeFixtures(t *testing.T) {
 	t.Parallel()
 

pkg/js/libs/mysql/mysql_private.go

@@ -76,24 +76,57 @@ func BuildDSN(opts MySQLOptions) (string, error) {
 	return dsn.String(), nil
 }
 
-// sandboxDSN enforces the local file access sandbox on a MySQL DSN. The
-// driver's allowAllFiles option lets a malicious server read any local file
-// off the host via LOAD DATA LOCAL INFILE, so it is only honored when -lfa is
-// enabled, mirroring the fs.ReadFile restriction.
-func sandboxDSN(dsn string, lfaAllowed bool) (string, error) {
+// sandboxDSN enforces local file and network policy on a MySQL DSN.
+// allowAllFiles is only honored when -lfa is enabled. tcp, tcp4 and tcp6
+// DSNs are rewritten to nucleitcp so the driver uses the fastdialer
+// registered in protocolstate. Unix sockets are treated as loopback for -lna.
+// Other networks are rejected.
+func sandboxDSN(executionId, dsn string) (string, error) {
 	cfg, err := mysql.ParseDSN(dsn)
 	if err != nil {
 		return "", err
 	}
-	if cfg.AllowAllFiles && !lfaAllowed {
+	opts := &types.Options{ExecutionId: executionId}
+	if cfg.AllowAllFiles && !protocolstate.IsLfaAllowed(opts) {
 		cfg.AllowAllFiles = false
 	}
+
+	netName := cfg.Net
+	if netName == "" {
+		netName = "tcp"
+	}
+	switch netName {
+	case "unix":
+		// IP-based policy never sees a socket path; gate on loopback like ldapi.
+		const unixPolicyHost = "127.0.0.1"
+		if !protocolstate.IsHostAllowed(executionId, unixPolicyHost) {
+			return "", protocolstate.ErrHostDenied.Msgf(unixPolicyHost)
+		}
+	case "tcp", "tcp4", "tcp6", "nucleitcp":
+		// Only nucleitcp has a registered dialer; the others would use the
+		// driver's plain net.Dialer and skip the dial-time policy check.
+		cfg.Net = "nucleitcp"
+		host := dsnPolicyHost(cfg.Addr)
+		if !protocolstate.IsHostAllowed(executionId, host) {
+			return "", protocolstate.ErrHostDenied.Msgf(host)
+		}
+	default:
+		return "", fmt.Errorf("unsupported mysql network %q", netName)
+	}
 	return cfg.FormatDSN(), nil
 }
 
+func dsnPolicyHost(addr string) string {
+	host, _, err := net.SplitHostPort(addr)
+	if err == nil && host != "" {
+		return host
+	}
+	return addr
+}
+
 // openDB opens a sandboxed MySQL connection from dsn.
 func openDB(executionId, dsn string) (*sql.DB, error) {
-	dsn, err := sandboxDSN(dsn, protocolstate.IsLfaAllowed(&types.Options{ExecutionId: executionId}))
+	dsn, err := sandboxDSN(executionId, dsn)
 	if err != nil {
 		return nil, err
 	}

pkg/js/libs/mysql/mysql_private_test.go

@@ -5,11 +5,14 @@ import (
 
 	"github.com/go-sql-driver/mysql"
 	"github.com/stretchr/testify/require"
+
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/protocolstate"
+	"github.com/projectdiscovery/nuclei/v3/pkg/types"
 )
 
 func TestSandboxDSN(t *testing.T) {
 	t.Run("strips allowAllFiles when lfa disabled", func(t *testing.T) {
-		got, err := sandboxDSN("root:x@nucleitcp(127.0.0.1:3306)/?allowAllFiles=true", false)
+		got, err := sandboxDSN("", "root:x@nucleitcp(127.0.0.1:3306)/?allowAllFiles=true")
 		require.NoError(t, err)
 
 		cfg, err := mysql.ParseDSN(got)
@@ -18,7 +21,14 @@ func TestSandboxDSN(t *testing.T) {
 	})
 
 	t.Run("keeps allowAllFiles when lfa enabled", func(t *testing.T) {
-		got, err := sandboxDSN("root:x@nucleitcp(127.0.0.1:3306)/?allowAllFiles=true", true)
+		executionID := "mysql-lfa-" + t.Name()
+		require.NoError(t, protocolstate.Init(&types.Options{
+			ExecutionId:          executionID,
+			AllowLocalFileAccess: true,
+		}))
+		t.Cleanup(func() { protocolstate.Close(executionID) })
+
+		got, err := sandboxDSN(executionID, "root:x@nucleitcp(127.0.0.1:3306)/?allowAllFiles=true")
 		require.NoError(t, err)
 
 		cfg, err := mysql.ParseDSN(got)
@@ -27,7 +37,7 @@ func TestSandboxDSN(t *testing.T) {
 	})
 
 	t.Run("leaves dsn without allowAllFiles untouched", func(t *testing.T) {
-		got, err := sandboxDSN("root:x@nucleitcp(127.0.0.1:3306)/", false)
+		got, err := sandboxDSN("", "root:x@nucleitcp(127.0.0.1:3306)/")
 		require.NoError(t, err)
 
 		cfg, err := mysql.ParseDSN(got)
@@ -35,8 +45,80 @@ func TestSandboxDSN(t *testing.T) {
 		require.False(t, cfg.AllowAllFiles)
 	})
 
+	t.Run("rewrites tcp to nucleitcp", func(t *testing.T) {
+		got, err := sandboxDSN("", "root:x@tcp(acme.com:3306)/")
+		require.NoError(t, err)
+
+		cfg, err := mysql.ParseDSN(got)
+		require.NoError(t, err)
+		require.Equal(t, "nucleitcp", cfg.Net)
+		require.Equal(t, "acme.com:3306", cfg.Addr)
+	})
+
+	t.Run("rewrites tcp4 and tcp6 to nucleitcp", func(t *testing.T) {
+		for _, netName := range []string{"tcp4", "tcp6"} {
+			got, err := sandboxDSN("", "root:x@"+netName+"(acme.com:3306)/")
+			require.NoError(t, err)
+
+			cfg, err := mysql.ParseDSN(got)
+			require.NoError(t, err)
+			require.Equal(t, "nucleitcp", cfg.Net)
+			require.Equal(t, "acme.com:3306", cfg.Addr)
+		}
+	})
+
+	t.Run("rejects unsupported network", func(t *testing.T) {
+		_, err := sandboxDSN("", "root:x@udp(acme.com:3306)/")
+		require.Error(t, err)
+		require.Contains(t, err.Error(), "udp")
+	})
+
 	t.Run("errors on invalid dsn", func(t *testing.T) {
-		_, err := sandboxDSN("::not-a-dsn::", false)
+		_, err := sandboxDSN("", "::not-a-dsn::")
 		require.Error(t, err)
 	})
+
+	t.Run("denies loopback tcp when lna enabled", func(t *testing.T) {
+		executionID := "mysql-lna-tcp-" + t.Name()
+		require.NoError(t, protocolstate.Init(&types.Options{
+			ExecutionId:                executionID,
+			RestrictLocalNetworkAccess: true,
+		}))
+		t.Cleanup(func() { protocolstate.Close(executionID) })
+
+		_, err := sandboxDSN(executionID, "vuser:vpass@tcp(127.0.0.1:3306)/vdb")
+		require.Error(t, err)
+		require.Contains(t, err.Error(), "127.0.0.1")
+	})
+
+	t.Run("denies loopback variants when lna enabled", func(t *testing.T) {
+		executionID := "mysql-lna-variants-" + t.Name()
+		require.NoError(t, protocolstate.Init(&types.Options{
+			ExecutionId:                executionID,
+			RestrictLocalNetworkAccess: true,
+		}))
+		t.Cleanup(func() { protocolstate.Close(executionID) })
+
+		for _, dsn := range []string{
+			"vuser:vpass@tcp4(127.0.0.1:3306)/vdb",
+			"vuser:vpass@tcp6([::1]:3306)/vdb",
+			"vuser:vpass@tcp([::ffff:127.0.0.1]:3306)/vdb",
+		} {
+			_, err := sandboxDSN(executionID, dsn)
+			require.Error(t, err, dsn)
+		}
+	})
+
+	t.Run("denies unix socket when lna enabled", func(t *testing.T) {
+		executionID := "mysql-lna-unix-" + t.Name()
+		require.NoError(t, protocolstate.Init(&types.Options{
+			ExecutionId:                executionID,
+			RestrictLocalNetworkAccess: true,
+		}))
+		t.Cleanup(func() { protocolstate.Close(executionID) })
+
+		_, err := sandboxDSN(executionID, "vuser:vpass@unix(/tmp/nuclei-mysql-lna.sock)/vdb")
+		require.Error(t, err)
+		require.Contains(t, err.Error(), "127.0.0.1")
+	})
 }

pkg/js/libs/redis/memo.redis.go

@@ -3,7 +3,6 @@ package redis
 
 import (
 	"context"
-
 	"errors"
 
 	"fmt"

pkg/js/libs/redis/redis.go

@@ -3,13 +3,12 @@ package redis
 import (
 	"bytes"
 	"context"
-	"errors"
 	"fmt"
-	"io"
 	"net"
 	"strconv"
 	"time"
 
+	"github.com/projectdiscovery/gologger"
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/protocolstate"
 	"github.com/redis/go-redis/v9"
 
@@ -269,9 +268,10 @@ var errReadLimit = fmt.Errorf("redis: reply exceeds %d bytes", maxReadBytes)
 // maxReadBytes.
 func newClient(addr, password string) *redis.Client {
 	opts := &redis.Options{
-		Addr:     addr,
-		Password: password,
-		DB:       0,
+		Addr:          addr,
+		Password:      password,
+		DB:            0,
+		DialerRetries: dialAttempts,
 	}
 	dial := redis.NewDialer(opts)
 	opts.Dialer = func(ctx context.Context, network, addr string) (net.Conn, error) {
@@ -285,138 +285,134 @@ func newClient(addr, password string) *redis.Client {
 	return redis.NewClient(opts)
 }
 
-// limitedConn fails a read once remaining is spent, and withholds a RESP length line whose
-// declared size is over the cap so go-redis never allocates it.
+// maxHeaderLen fits a type byte, any int64, and "\r". Redis never sends a longer length header.
+const maxHeaderLen = 32
+
+// limitedConn fails a read once remaining is spent, and ends a read before the "\n" of a RESP
+// length header whose declared size is over the cap, so go-redis never gets the line to allocate
+// from.
+//
+// It follows the reply's framing as the bytes pass: a bulk payload is skipped by its declared
+// length, so only header lines are checked. A payload line such as "*abc" is data.
 type limitedConn struct {
 	net.Conn
 	remaining int64
-	pending   []byte
-	partial   []byte
-	err       error
-	scratch   [32 << 10]byte
+	// payload is how many bytes of the current bulk payload, "\r\n" included, are still to come.
+	payload int64
+	// header is the start of the current header line, without its "\n".
+	header   []byte
+	overlong bool
+	err      error
 }
 
 func (c *limitedConn) Read(b []byte) (int, error) {
-	if len(b) == 0 {
+	if c.err != nil {
+		return 0, c.err
+	}
+	if c.remaining <= 0 {
+		return 0, errReadLimit
+	}
+	if int64(len(b)) > c.remaining {
+		b = b[:c.remaining]
+	}
+	n, err := c.Conn.Read(b)
+	c.remaining -= int64(n)
+	// Errors from the connection, deadlines included, are returned as they come and not kept: a
+	// read deadline belongs to one call. go-redis peeks for push notifications with a few
+	// microseconds of deadline, and keeping that timeout would fail the command that follows.
+	kept, scanErr := c.scan(b[:n])
+	if scanErr == nil {
+		return n, err
+	}
+	c.err = scanErr
+	if kept == 0 {
+		return 0, scanErr
+	}
+	return kept, nil
+}
+
+// scan follows the RESP framing of p and returns how many of its bytes may be returned.
+func (c *limitedConn) scan(p []byte) (int, error) {
+	for i := 0; i < len(p); {
+		if c.payload > 0 {
+			skip := min(c.payload, int64(len(p)-i))
+			c.payload -= skip
+			i += int(skip)
+			continue
+		}
+		end := bytes.IndexByte(p[i:], '\n')
+		if end < 0 {
+			c.keepHeader(p[i:])
+			return len(p), nil
+		}
+		c.keepHeader(p[i : i+end])
+		payload, err := declaredPayload(c.header, c.overlong)
+		c.header, c.overlong = c.header[:0], false
+		if err != nil {
+			return i + end, err
+		}
+		c.payload = payload
+		i += end + 1
+	}
+	return len(p), nil
+}
+
+func (c *limitedConn) keepHeader(p []byte) {
+	if room := maxHeaderLen - len(c.header); len(p) > room {
+		p = p[:room]
+		c.overlong = true
+	}
+	c.header = append(c.header, p...)
+}
+
+// declaredPayload returns how many payload bytes, "\r\n" included, follow a RESP header line
+// given without its "\n". It returns errReadLimit when go-redis would allocate more than the cap
+// from the line: the declared byte length for bulk strings, one interface per declared element
+// for arrays, sets, maps, attributes, and pushes.
+func declaredPayload(line []byte, overlong bool) (int64, error) {
+	if len(line) == 0 {
 		return 0, nil
 	}
-	if len(c.pending) == 0 {
-		if err := c.fill(); err != nil {
-			return 0, err
-		}
-	}
-	n := copy(b, c.pending)
-	c.pending = c.pending[n:]
-	return n, nil
-}
-
-func (c *limitedConn) fill() error {
-	if len(c.pending) > 0 {
-		return nil
-	}
-	if c.err != nil {
-		return c.err
-	}
-	for len(c.pending) == 0 {
-		if c.remaining <= 0 {
-			c.err = errReadLimit
-			return c.err
-		}
-		want := int64(len(c.scratch))
-		if want > c.remaining {
-			want = c.remaining
-		}
-		n, err := c.Conn.Read(c.scratch[:want])
-		if n > 0 {
-			c.remaining -= int64(n)
-			if aerr := c.accept(c.scratch[:n]); aerr != nil {
-				c.err = aerr
-				if len(c.pending) > 0 {
-					return nil
-				}
-				return aerr
-			}
-		}
-		if err != nil {
-			// A read deadline belongs to one call. go-redis peeks for push notifications
-			// with a few microseconds of deadline, and on Windows that peek always runs.
-			// Keeping the timeout fails the command that follows.
-			if timeoutErr(err) {
-				if len(c.pending) > 0 {
-					return nil
-				}
-				return err
-			}
-			c.err = err
-			if len(c.pending) > 0 {
-				return nil
-			}
-			return err
-		}
-		if n == 0 {
-			c.err = io.ErrUnexpectedEOF
-			return c.err
-		}
-	}
-	return nil
-}
-
-func (c *limitedConn) accept(p []byte) error {
-	c.partial = append(c.partial, p...)
-	for {
-		i := bytes.IndexByte(c.partial, '\n')
-		if i < 0 {
-			if int64(len(c.partial)) > maxReadBytes {
-				c.partial = nil
-				return errReadLimit
-			}
-			return nil
-		}
-		line := c.partial[:i+1]
-		if lengthTooLarge(line) {
-			c.partial = nil
-			return errReadLimit
-		}
-		copied := make([]byte, len(line))
-		copy(copied, line)
-		c.pending = append(c.pending, copied...)
-		rest := make([]byte, len(c.partial)-(i+1))
-		copy(rest, c.partial[i+1:])
-		c.partial = rest
-	}
-}
-
-// lengthTooLarge reports whether line is a RESP length header go-redis would allocate from.
-// Bulk strings allocate the declared byte length. Arrays, sets, maps, attributes, and pushes
-// allocate one interface per declared element.
-func lengthTooLarge(line []byte) bool {
-	if len(line) < 3 {
-		return false
-	}
 	var limit int64
+	bulk := false
 	switch line[0] {
 	case '$', '=', '!':
-		limit = maxReadBytes
+		limit, bulk = maxReadBytes, true
 	case '*', '~', '>', '%', '|':
 		limit = maxAggregate
 	default:
-		return false
+		return 0, nil
 	}
-	body := line[1 : len(line)-1]
-	if len(body) > 0 && body[len(body)-1] == '\r' {
-		body = body[:len(body)-1]
+	if overlong {
+		return 0, errReadLimit
 	}
+	body := bytes.TrimSuffix(line[1:], []byte("\r"))
 	if len(body) == 0 || body[0] == '-' {
-		return false
+		return 0, nil
 	}
 	n, err := strconv.ParseInt(string(body), 10, 64)
-	if err != nil {
-		return true
+	if err != nil || n > limit {
+		return 0, errReadLimit
 	}
-	return n > limit
+	if !bulk {
+		return 0, nil
+	}
+	return n + 2, nil
 }
 
-func timeoutErr(err error) bool {
-	var netErr net.Error
-	return errors.As(err, &netErr) && netErr.Timeout()
+// dialAttempts replaces go-redis's default of 5. The extra attempts run in a goroutine that
+// ignores the caller's context, so they kept dialing a filtered port for ~25s after the
+// template had given up on it.
+const dialAttempts = 1
+
+func init() {
+	redis.SetLogger(warningLogger{})
+}
+
+// warningLogger hands go-redis's logs to gologger as warnings. Its default logger writes them
+// straight to stderr, where a failed dial against the scanned host reads as an error.
+type warningLogger struct{}
+
+func (warningLogger) Printf(_ context.Context, format string, v ...interface{}) {
+	gologger.Warning().Msgf(format, v...)
 }

pkg/js/libs/redis/redis_dial_test.go

@@ -0,0 +1,81 @@
+package redis
+
+import (
+	"context"
+	"net"
+	"runtime"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/projectdiscovery/gologger"
+	"github.com/projectdiscovery/gologger/levels"
+	"github.com/projectdiscovery/gologger/writer"
+	"github.com/stretchr/testify/require"
+)
+
+// go-redis retries a failed dial 4 times, 100ms apart, on each of a command's 4 attempts: a closed
+// port took at least 1.6s to fail.
+func TestGetServerInfoFailsFastOnAClosedPort(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("windows takes about a second to report a refused localhost dial")
+	}
+
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	addr, ok := listener.Addr().(*net.TCPAddr)
+	require.True(t, ok)
+	require.NoError(t, listener.Close())
+
+	start := time.Now()
+	_, err = getServerInfo(context.Background(), "redis-closed-port", "127.0.0.1", addr.Port)
+	require.Error(t, err)
+	require.Less(t, time.Since(start), time.Second)
+}
+
+// go-redis's default logger wrote each failed dial straight to stderr, past nuclei's logger and
+// its level.
+func TestGoRedisLogsAreGologgerWarnings(t *testing.T) {
+	logs := &recordingWriter{}
+	gologger.DefaultLogger.SetWriter(logs)
+	gologger.DefaultLogger.SetMaxLevel(levels.LevelWarning)
+	t.Cleanup(func() {
+		// gologger's own defaults: it has no getter to save the previous ones.
+		gologger.DefaultLogger.SetWriter(writer.NewCLI())
+		gologger.DefaultLogger.SetMaxLevel(levels.LevelInfo)
+	})
+
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	addr, ok := listener.Addr().(*net.TCPAddr)
+	require.True(t, ok)
+	require.NoError(t, listener.Close())
+
+	_, err = getServerInfo(context.Background(), "redis-log-level", "127.0.0.1", addr.Port)
+	require.Error(t, err)
+
+	warnings := logs.at(levels.LevelWarning)
+	require.NotEmpty(t, warnings)
+	require.Contains(t, warnings[0], "failed to dial")
+}
+
+type recordingWriter struct {
+	mu    sync.Mutex
+	lines map[levels.Level][]string
+}
+
+func (w *recordingWriter) Write(data []byte, level levels.Level) {
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	if w.lines == nil {
+		w.lines = make(map[levels.Level][]string)
+	}
+	w.lines[level] = append(w.lines[level], strings.TrimSpace(string(data)))
+}
+
+func (w *recordingWriter) at(level levels.Level) []string {
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	return w.lines[level]
+}

pkg/js/libs/redis/redis_test.go

@@ -5,6 +5,7 @@ import (
 	"bytes"
 	"context"
 	"fmt"
+	"io"
 	"net"
 	"runtime"
 	"strconv"
@@ -90,37 +91,58 @@ func TestGetServerInfoAuthBoundsAReplyWithoutNewline(t *testing.T) {
 }
 
 func TestGetServerInfoAuthReadsAnInfoReply(t *testing.T) {
-	listener, err := net.Listen("tcp", "127.0.0.1:0")
-	require.NoError(t, err)
-	t.Cleanup(func() {
-		_ = listener.Close()
-	})
+	tests := []struct {
+		name string
+		info string
+	}{
+		{name: "info", info: "# Server\r\nredis_version:7.2.4\r\n"},
+		// A bulk payload's lines are data, even when they start like a RESP length header.
+		{name: "header-like lines", info: "# Server\r\n*abc\r\n$ 5 dollars\r\n%x\r\n$99999999999\r\n"},
+		// Each line used to cost a copy of what was left of the read buffer.
+		{name: "many short lines", info: strings.Repeat("\n", 1<<20)},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			listener, err := net.Listen("tcp", "127.0.0.1:0")
+			require.NoError(t, err)
+			t.Cleanup(func() {
+				_ = listener.Close()
+			})
 
-	const info = "# Server\r\nredis_version:7.2.4\r\n"
-	go func() {
-		for {
-			conn, err := listener.Accept()
-			if err != nil {
-				return
-			}
-			go serveRESP(conn, info)
-		}
-	}()
+			go func() {
+				for {
+					conn, err := listener.Accept()
+					if err != nil {
+						return
+					}
+					go serveRESP(conn, tt.info)
+				}
+			}()
 
-	executionID := "redis-" + strings.ReplaceAll(t.Name(), "/", "-")
-	require.NoError(t, protocolstate.Init(&types.Options{ExecutionId: executionID}))
-	t.Cleanup(func() {
-		protocolstate.Close(executionID)
-	})
+			executionID := "redis-" + strings.ReplaceAll(t.Name(), "/", "-")
+			require.NoError(t, protocolstate.Init(&types.Options{ExecutionId: executionID}))
+			t.Cleanup(func() {
+				protocolstate.Close(executionID)
+			})
 
-	addr, ok := listener.Addr().(*net.TCPAddr)
-	require.True(t, ok)
+			addr, ok := listener.Addr().(*net.TCPAddr)
+			require.True(t, ok)
 
-	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
-	defer cancel()
-	got, err := getServerInfoAuth(ctx, executionID, "127.0.0.1", addr.Port, "")
-	require.NoError(t, err)
-	require.Equal(t, info, got)
+			var before runtime.MemStats
+			runtime.ReadMemStats(&before)
+
+			ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+			defer cancel()
+			got, err := getServerInfoAuth(ctx, executionID, "127.0.0.1", addr.Port, "")
+
+			var after runtime.MemStats
+			runtime.ReadMemStats(&after)
+			require.NoError(t, err)
+			require.Equal(t, tt.info, got)
+			growth := int64(after.TotalAlloc - before.TotalAlloc)
+			require.Less(t, growth, int64(64<<20), "allocated %d MiB reading one reply", growth>>20)
+		})
+	}
 }
 
 // serveRESP answers HELLO as a Redis without RESP3 does, so the client falls back to RESP2, and
@@ -182,26 +204,85 @@ func readRESPCommand(rd *bufio.Reader) (string, error) {
 	return command, nil
 }
 
-func TestLengthTooLarge(t *testing.T) {
+func TestDeclaredPayload(t *testing.T) {
 	tests := []struct {
-		line string
-		want bool
+		line     string
+		overlong bool
+		payload  int64
+		tooLarge bool
 	}{
-		{line: fmt.Sprintf("$%d\r\n", maxReadBytes), want: false},
-		{line: fmt.Sprintf("$%d\r\n", maxReadBytes+1), want: true},
-		{line: "$-1\r\n", want: false},
-		{line: "+OK\r\n", want: false},
-		{line: fmt.Sprintf("*%d\r\n", maxAggregate), want: false},
-		{line: fmt.Sprintf("*%d\r\n", maxAggregate+1), want: true},
-		{line: "$999999999999999999999\r\n", want: true},
+		{line: fmt.Sprintf("$%d\r", maxReadBytes), payload: maxReadBytes + 2},
+		{line: fmt.Sprintf("$%d\r", maxReadBytes+1), tooLarge: true},
+		{line: "=10\r", payload: 12},
+		{line: "$-1\r"},
+		{line: "+OK\r"},
+		{line: "+OK\r", overlong: true},
+		{line: fmt.Sprintf("*%d\r", maxAggregate)},
+		{line: fmt.Sprintf("*%d\r", maxAggregate+1), tooLarge: true},
+		{line: "$999999999999999999999\r", tooLarge: true},
+		{line: "*abc\r", tooLarge: true},
+		{line: "$000000000000000000000000000000", overlong: true, tooLarge: true},
 	}
 	for _, tt := range tests {
 		t.Run(tt.line, func(t *testing.T) {
-			require.Equal(t, tt.want, lengthTooLarge([]byte(tt.line)))
+			payload, err := declaredPayload([]byte(tt.line), tt.overlong)
+			if tt.tooLarge {
+				require.ErrorIs(t, err, errReadLimit)
+				return
+			}
+			require.NoError(t, err)
+			require.Equal(t, tt.payload, payload)
 		})
 	}
 }
 
+// The check that rejects a length header has to land on the "\n" when the header arrives over
+// several reads, and payload bytes have to pass whatever they contain.
+func TestLimitedConnFollowsFramingAcrossReads(t *testing.T) {
+	const ok = "*2\r\n$5\r\n*abc\n\r\n$-1\r\n+OK\r\n"
+	tooLarge := fmt.Sprintf("$%d\r\n", maxReadBytes+1)
+
+	tests := []struct {
+		name    string
+		stream  string
+		want    string
+		wantErr error
+	}{
+		{name: "payload passes", stream: ok, want: ok},
+		{name: "header rejected", stream: ok + tooLarge + "x", want: ok + tooLarge[:len(tooLarge)-1], wantErr: errReadLimit},
+	}
+	for _, tt := range tests {
+		for _, size := range []int{1, 3, 64} {
+			t.Run(fmt.Sprintf("%s/%d", tt.name, size), func(t *testing.T) {
+				lc := &limitedConn{Conn: &chunkedConn{data: []byte(tt.stream), size: size}, remaining: maxReadBytes}
+				got, err := io.ReadAll(lc)
+				require.Equal(t, tt.want, string(got))
+				if tt.wantErr != nil {
+					require.ErrorIs(t, err, tt.wantErr)
+					return
+				}
+				require.NoError(t, err)
+			})
+		}
+	}
+}
+
+// chunkedConn returns data at most size bytes per read, then io.EOF.
+type chunkedConn struct {
+	net.Conn
+	data []byte
+	size int
+}
+
+func (c *chunkedConn) Read(b []byte) (int, error) {
+	if len(c.data) == 0 {
+		return 0, io.EOF
+	}
+	n := copy(b[:min(len(b), c.size)], c.data)
+	c.data = c.data[n:]
+	return n, nil
+}
+
 // A short RESP length header used to make go-redis allocate the declared size before reading
 // any of the payload, so the byte cap never applied.
 func TestGetServerInfoAuthRejectsAHugeDeclaredLength(t *testing.T) {

pkg/loader/workflow/workflow_loader.go

@@ -16,6 +16,8 @@ import (
 type workflowLoader struct {
 	pathFilter *filter.PathFilter
 	tagFilter  *templates.TagFilter
+	// the exclusions of tagFilter other than severities, see filters
+	exclusions *templates.TagFilter
 	options    *protocols.ExecutorOptions
 
 	// every workflow step that selects templates by tag searches the whole
@@ -45,12 +47,29 @@ func NewLoader(options *protocols.ExecutorOptions) (model.WorkflowLoader, error)
 	if err != nil {
 		return nil, err
 	}
+	// IncludeTags is not an inclusion filter: it lifts the exclusion of the
+	// same tags
+	exclusions, err := templates.NewTagFilter(&templates.TagFilterConfig{
+		ExcludeTags:      options.Options.ExcludeTags,
+		IncludeTags:      options.Options.IncludeTags,
+		ExcludeIds:       options.Options.ExcludeIds,
+		ExcludeProtocols: options.Options.ExcludeProtocols,
+	})
+	if err != nil {
+		return nil, err
+	}
 	pathFilter := filter.NewPathFilter(&filter.PathFilterConfig{
 		IncludedTemplates: options.Options.IncludeTemplates,
 		ExcludedTemplates: options.Options.ExcludedTemplates,
 	}, options.Catalog)
 
-	loader := &workflowLoader{pathFilter: pathFilter, tagFilter: tagFilter, options: options, parsed: make(map[string]*templates.Template)}
+	loader := &workflowLoader{
+		pathFilter: pathFilter,
+		tagFilter:  tagFilter,
+		exclusions: exclusions,
+		options:    options,
+		parsed:     make(map[string]*templates.Template),
+	}
 	loader.templatesDirectoryPaths = sync.OnceValue(loader.listTemplatesDirectory)
 	return loader, nil
 }
@@ -85,20 +104,34 @@ func (w *workflowLoader) parseTemplate(path string) *templates.Template {
 	return template
 }
 
-func (w *workflowLoader) GetTemplatePathsByTags(templateTags []string) []string {
+func (w *workflowLoader) GetTemplatePathsByTags(templateTags []string, noValidate bool) []string {
 	var loadedTemplates []string
 	for _, path := range w.templatesDirectoryPaths() {
+		// the step's tags select the templates and the filters then decide
+		// which of them load, as for the templates of a template step
 		template := w.parseTemplate(path)
-		if template == nil {
+		if template == nil || !selectedByTags(template, templateTags) {
 			continue
 		}
-		if loaded, _ := templates.MatchLoadFilters(template, path, w.tagFilter, templateTags); loaded {
+		if matched, _ := templates.MatchLoadFilters(template, path, w.filters(noValidate), nil); matched {
 			loadedTemplates = append(loadedTemplates, path)
 		}
 	}
 	return loadedTemplates
 }
 
+// selectedByTags reports whether a workflow step that selects templates by
+// tags selects template, which it does when template has one of tags.
+// Workflows are never selected, as a workflow cannot run as a step.
+func selectedByTags(template *templates.Template, tags []string) bool {
+	if len(template.Workflows) > 0 {
+		return false
+	}
+	return slices.ContainsFunc(template.Info.Tags.ToSlice(), func(tag string) bool {
+		return slices.Contains(tags, tag)
+	})
+}
+
 func (w *workflowLoader) GetTemplatePaths(templatesList []string, noValidate bool) []string {
 	includedTemplates, errs := w.options.Catalog.GetTemplatesPath(templatesList)
 	for template, err := range errs {
@@ -108,12 +141,27 @@ func (w *workflowLoader) GetTemplatePaths(templatesList []string, noValidate boo
 
 	loadedTemplates := make([]string, 0, len(templatesPathMap))
 	for templatePath := range templatesPathMap {
-		matched, err := w.options.Parser.LoadTemplate(templatePath, w.tagFilter, nil, w.options.Catalog)
+		matched, err := w.options.Parser.LoadTemplate(templatePath, w.filters(noValidate), nil, w.options.Catalog)
 		if err != nil && !matched {
 			gologger.Warning().Msg(err.Error())
-		} else if matched || noValidate {
+		} else if matched {
 			loadedTemplates = append(loadedTemplates, templatePath)
 		}
 	}
 	return loadedTemplates
 }
+
+// filters returns the filters that decide which templates a workflow step
+// loads. With noValidate, set for a step whose subtemplates loaded, only the
+// exclusions apply: the step must load for its subtemplates to run, but an
+// excluded template never runs. Excluded severities do not apply either: a
+// step with subtemplates does not report its own results, so excluding its
+// severity would only keep its subtemplates from running. A late
+// out-of-band (interactsh) match of the step is the exception: it is
+// reported whatever its severity, as with -s.
+func (w *workflowLoader) filters(noValidate bool) *templates.TagFilter {
+	if noValidate {
+		return w.exclusions
+	}
+	return w.tagFilter
+}

pkg/loader/workflow/workflow_loader_test.go

@@ -1,6 +1,8 @@
 package workflow
 
 import (
+	"maps"
+	"slices"
 	"testing"
 
 	"github.com/projectdiscovery/nuclei/v3/pkg/catalog"
@@ -8,8 +10,12 @@ import (
 	"github.com/projectdiscovery/nuclei/v3/pkg/model/types/severity"
 	"github.com/projectdiscovery/nuclei/v3/pkg/model/types/stringslice"
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols"
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/dns"
 	"github.com/projectdiscovery/nuclei/v3/pkg/templates"
+	templateTypes "github.com/projectdiscovery/nuclei/v3/pkg/templates/types"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/projectdiscovery/nuclei/v3/pkg/utils/stats"
+	"github.com/projectdiscovery/nuclei/v3/pkg/workflows"
 	"github.com/stretchr/testify/require"
 )
 
@@ -63,9 +69,87 @@ func TestGetTemplatePathsByTagsReadsTemplatesDirectoryOnce(t *testing.T) {
 	require.NoError(t, err)
 	walksBeforeLookups := cat.walks
 
-	require.Equal(t, []string{"wordpress.yaml"}, loader.GetTemplatePathsByTags([]string{"wordpress"}))
-	require.Equal(t, []string{"jira.yaml"}, loader.GetTemplatePathsByTags([]string{"jira"}))
+	require.Equal(t, []string{"wordpress.yaml"}, loader.GetTemplatePathsByTags([]string{"wordpress"}, false))
+	require.Equal(t, []string{"jira.yaml"}, loader.GetTemplatePathsByTags([]string{"jira"}, false))
 
 	require.Equal(t, 1, cat.walks-walksBeforeLookups, "templates directory walked per lookup")
 	require.Equal(t, map[string]int{"wordpress.yaml": 1, "jira.yaml": 1}, parser.parses, "templates parsed per lookup")
 }
+
+func TestGetTemplatePathsByTagsNoValidateAppliesExclusionsButSeverity(t *testing.T) {
+	template := func(id string, tags ...string) *templates.Template {
+		return &templates.Template{ID: id, Info: model.Info{
+			Name:           id,
+			Authors:        stringslice.StringSlice{Value: "author"},
+			Tags:           stringslice.StringSlice{Value: tags},
+			SeverityHolder: severity.Holder{Severity: severity.Info},
+		}}
+	}
+	excludedSeverity := template("excluded-severity", "step")
+	excludedSeverity.Info.SeverityHolder.Severity = severity.Low
+	excludedProtocol := template("excluded-protocol", "step")
+	excludedProtocol.RequestsDNS = []*dns.Request{{}}
+	workflow := template("workflow", "step")
+	workflow.Workflows = []*workflows.WorkflowTemplate{{Template: "selected.yaml"}}
+	parser := &countingParser{
+		templates: map[string]*templates.Template{
+			"selected.yaml":          template("selected", "step"),
+			"included.yaml":          template("included", "step", "included"),
+			"excluded-tag.yaml":      template("excluded-tag", "step", "excluded"),
+			"excluded-id.yaml":       template("excluded-id", "step"),
+			"excluded-severity.yaml": excludedSeverity,
+			"excluded-protocol.yaml": excludedProtocol,
+			"untagged.yaml":          template("untagged", "other"),
+			"workflow.yaml":          workflow,
+		},
+		parses: make(map[string]int),
+	}
+	options := types.DefaultOptions()
+	options.IncludeIds = []string{"untagged"}
+	options.ExcludeTags = []string{"excluded", "included"}
+	options.IncludeTags = []string{"included"}
+	options.ExcludeIds = []string{"excluded-id"}
+	options.ExcludeSeverities = severity.Severities{severity.Low}
+	options.ExcludeProtocols = templateTypes.ProtocolTypes{templateTypes.DNSProtocol}
+	cat := &countingCatalog{paths: slices.Collect(maps.Keys(parser.templates))}
+	loader, err := NewLoader(&protocols.ExecutorOptions{Options: options, Catalog: cat, Parser: parser})
+	require.NoError(t, err)
+
+	require.Empty(t, loader.GetTemplatePathsByTags([]string{"step"}, false))
+	loaded := loader.GetTemplatePathsByTags([]string{"step"}, true)
+	require.ElementsMatch(t, []string{"selected.yaml", "included.yaml", "excluded-severity.yaml"}, loaded)
+}
+
+func TestGetTemplatePathsByTagsCountsSyntaxWarningsOfLoadedTemplates(t *testing.T) {
+	// a template without severity has a syntax warning
+	unrated := &templates.Template{ID: "unrated", Info: model.Info{
+		Name:    "unrated",
+		Authors: stringslice.StringSlice{Value: "author"},
+		Tags:    stringslice.StringSlice{Value: "step"},
+	}}
+	tests := []struct {
+		name     string
+		ids      []string
+		warnings int64
+	}{
+		{name: "loaded", warnings: 1},
+		{name: "filtered out", ids: []string{"other"}, warnings: 0},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			parser := &countingParser{
+				templates: map[string]*templates.Template{"unrated.yaml": unrated},
+				parses:    make(map[string]int),
+			}
+			options := types.DefaultOptions()
+			options.IncludeIds = tc.ids
+			cat := &countingCatalog{paths: []string{"unrated.yaml"}}
+			loader, err := NewLoader(&protocols.ExecutorOptions{Options: options, Catalog: cat, Parser: parser})
+			require.NoError(t, err)
+
+			before := stats.GetValue(templates.TemplateSyntaxWarningStats)
+			loader.GetTemplatePathsByTags([]string{"step"}, false)
+			require.Equal(t, tc.warnings, stats.GetValue(templates.TemplateSyntaxWarningStats)-before)
+		})
+	}
+}

pkg/model/model.go

@@ -1,9 +1,15 @@
 package model
 
 import (
+	"slices"
+	"strings"
+
 	"github.com/invopop/jsonschema"
 	"github.com/projectdiscovery/nuclei/v3/pkg/model/types/severity"
 	"github.com/projectdiscovery/nuclei/v3/pkg/model/types/stringslice"
+	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/common/marker"
+	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/projectdiscovery/nuclei/v3/pkg/utils/json"
 )
 
 type schemaMetadata struct {
@@ -47,15 +53,28 @@ type Info struct {
 	//
 	//   You can go in-depth here on what the template actually does.
 	//
+	//   Description, impact and remediation may hold {{placeholders}} filled for
+	//   each finding: named extractors, payloads, variables and DSL helpers
+	//   (network helpers excluded), and the result fields host, port, ip, url,
+	//   scheme, path, matched-at, matcher-name and extractor-name, which win over
+	//   values of the same name. Use join(", ", name) for an extractor with
+	//   several values, and brackets for a hyphenated name inside a helper:
+	//   to_upper([matcher-name]). Unknown placeholders stay as written, and so
+	//   does a field whose expression fails. Global matcher templates match on
+	//   other templates' responses, so only the result fields fill their text.
+	//
 	// examples:
 	//   - value: "\"Bower is a package manager which stores package information in the bower.json file\""
 	//   - value: "\"Subversion ALM for the enterprise before 8.8.2 allows reflected XSS at multiple locations\""
+	//   - value: "\"Redis {{version}} on {{host}}:{{port}} accepts the default password\""
 	Description string `json:"description,omitempty" yaml:"description,omitempty" jsonschema:"title=description of the template,description=In-depth explanation on what the template does,type=string,example=Bower is a package manager which stores package information in the bower.json file"`
 	// description: |
 	//   Impact of the template.
 	//
 	//   You can go in-depth here on impact of the template.
 	//
+	//   Supports the same placeholders as description.
+	//
 	// examples:
 	//   - value: "\"Successful exploitation of this vulnerability could allow an attacker to execute arbitrary SQL queries, potentially leading to unauthorized access, data leakage, or data manipulation.\""
 	//   - value: "\"Successful exploitation of this vulnerability could allow an attacker to execute arbitrary script code in the context of the victim's browser, potentially leading to session hijacking, defacement, or theft of sensitive information.\""
@@ -89,11 +108,70 @@ type Info struct {
 	//
 	//   You can go in-depth here on how to mitigate the problem found by this template.
 	//
+	//   Supports the same placeholders as description.
+	//
 	// examples:
 	//   - value: "\"Change the default administrative username and password of Apache ActiveMQ by editing the file jetty-realm.properties\""
 	Remediation string `json:"remediation,omitempty" yaml:"remediation,omitempty" jsonschema:"title=remediation steps for the template,description=In-depth explanation on how to fix the issues found by the template,example=Change the default administrative username and password of Apache ActiveMQ by editing the file jetty-realm.properties,type=string"`
 }
 
+// MarshalJSON encodes the info with every map nested in Metadata keyed by strings.
+func (i Info) MarshalJSON() ([]byte, error) {
+	// info has the fields of Info but not this method, so marshaling it does not recurse.
+	type info Info
+	encoded := info(i)
+	metadata, err := types.ToJSONMap(i.Metadata)
+	if err != nil {
+		return nil, err
+	}
+	encoded.Metadata = metadata
+	return json.Marshal(encoded)
+}
+
+// TextField is a free-text field of the info, labeled as reports show it.
+type TextField struct {
+	Label string
+	Value *string
+}
+
+// PlaceholderFields returns the free-text fields that may hold {{placeholders}}
+// filled for each finding. Rendering, redaction and the sarif export all walk
+// this one list, so a field added here is handled by every one of them.
+func (i *Info) PlaceholderFields() []TextField {
+	return []TextField{
+		{Label: "Description", Value: &i.Description},
+		{Label: "Impact", Value: &i.Impact},
+		{Label: "Remediation", Value: &i.Remediation},
+	}
+}
+
+// HasPlaceholders reports whether the field holds a {{...}} or §...§ marker
+// the renderer could fill.
+func (f TextField) HasPlaceholders() bool {
+	return strings.Contains(*f.Value, marker.ParenthesisOpen) || strings.Contains(*f.Value, marker.General)
+}
+
+// HasPlaceholders reports whether any placeholder field holds a marker.
+func (i *Info) HasPlaceholders() bool {
+	return slices.ContainsFunc(i.PlaceholderFields(), TextField.HasPlaceholders)
+}
+
+// FilledFields returns the placeholder fields whose text differs from written,
+// the info as the template wrote it. A nil written means nothing was filled.
+func (i *Info) FilledFields(written *Info) []TextField {
+	if written == nil {
+		return nil
+	}
+	writtenFields := written.PlaceholderFields()
+	var filled []TextField
+	for n, field := range i.PlaceholderFields() {
+		if *field.Value != *writtenFields[n].Value {
+			filled = append(filled, field)
+		}
+	}
+	return filled
+}
+
 // JSONSchemaProperty returns the JSON schema property for the Info object.
 func (i Info) JSONSchemaExtend(base *jsonschema.Schema) {
 	// since we are re-using a stringslice and rawStringSlice everywhere, we can extend/edit the schema here

pkg/model/model_test.go

@@ -35,6 +35,34 @@ func TestInfoJsonMarshal(t *testing.T) {
 	require.Equal(t, expected, string(result))
 }
 
+func TestInfoJsonMarshalNestedYAMLMetadata(t *testing.T) {
+	var info Info
+	err := yaml.Unmarshal([]byte(`
+name: Nested metadata
+author: pdteam
+tags: misc
+severity: info
+metadata:
+  labels:
+    en: Hello
+    1: one
+  items:
+    - name: first
+      tags:
+        fr: Bonjour
+  plain: value
+`), &info)
+	require.NoError(t, err)
+	require.IsType(t, map[interface{}]interface{}{}, info.Metadata["labels"])
+
+	for _, value := range []interface{}{info, &info} {
+		result, err := json.Marshal(value)
+		require.NoError(t, err)
+		require.JSONEq(t, `{"name":"Nested metadata","author":["pdteam"],"tags":["misc"],"severity":"info","metadata":{"labels":{"en":"Hello","1":"one"},"items":[{"name":"first","tags":{"fr":"Bonjour"}}],"plain":"value"}}`, string(result))
+	}
+	require.IsType(t, map[interface{}]interface{}{}, info.Metadata["labels"], "marshaling must not modify the info")
+}
+
 func TestInfoYamlMarshal(t *testing.T) {
 	info := Info{
 		Name:           "Test Template Name",
@@ -135,3 +163,21 @@ func TestUnmarshal(t *testing.T) {
 	info2 := assertUnmarshalledTemplateInfo(t, yamlPayload2)
 	require.Equal(t, info1, info2)
 }
+
+func TestInfoFilledFields(t *testing.T) {
+	written := Info{Description: "{{host}} is exposed", Impact: "Data loss"}
+	filled := written
+	filled.Description = "example.com is exposed"
+
+	require.Nil(t, filled.FilledFields(nil))
+	fields := filled.FilledFields(&written)
+	require.Len(t, fields, 1)
+	require.Equal(t, "Description", fields[0].Label)
+	require.Equal(t, "example.com is exposed", *fields[0].Value)
+}
+
+func TestInfoHasPlaceholders(t *testing.T) {
+	require.False(t, (&Info{Description: "Redis is exposed"}).HasPlaceholders())
+	require.True(t, (&Info{Impact: "{{host}} is exposed"}).HasPlaceholders())
+	require.True(t, (&Info{Remediation: "rotate §password§"}).HasPlaceholders())
+}

pkg/model/workflow_loader.go

@@ -4,9 +4,13 @@ package model
 
 // WorkflowLoader is a loader interface required for workflow initialization.
 type WorkflowLoader interface {
-	// GetTemplatePathsByTags returns a list of template paths based on the provided tags from the templates directory
-	GetTemplatePathsByTags(tags []string) []string
+	// GetTemplatePathsByTags returns the paths of the templates in the
+	// templates directory that have one of tags. noValidate is as in
+	// GetTemplatePaths.
+	GetTemplatePathsByTags(tags []string, noValidate bool) []string
 
-	// GetTemplatePaths takes a list of templates and returns paths for them
+	// GetTemplatePaths takes a list of templates and returns paths for them.
+	// With noValidate, only the exclusion filters apply, except excluded
+	// severities: excluded tags, IDs, protocols and paths.
 	GetTemplatePaths(templatesList []string, noValidate bool) []string
 }

pkg/operators/common/dsl/network_policy_test.go

@@ -3,6 +3,8 @@ package dsl
 import (
 	"fmt"
 	"net"
+	"os"
+	"path/filepath"
 	"strings"
 	"sync"
 	"sync/atomic"
@@ -44,6 +46,12 @@ func networkTestDNS(t *testing.T, address string) (string, *atomic.Int32) {
 
 func networkTestOptions(t *testing.T, options *types.Options) *types.Options {
 	t.Helper()
+	// fastdialer adds the nameservers of the system resolver file before the
+	// stub resolvers and picks them round robin, so the stub answer would
+	// depend on the host. An empty resolver file keeps only the stub.
+	resolverFile := filepath.Join(t.TempDir(), "resolv.conf")
+	require.NoError(t, os.WriteFile(resolverFile, nil, 0o600))
+	t.Setenv("RESOLVERS_PATH", resolverFile)
 	if options.ExecutionId == "" {
 		options.ExecutionId = t.Name()
 	}

pkg/operators/common/llm/llm.go

@@ -21,6 +21,22 @@ type Client interface {
 	Complete(ctx context.Context, prompt string, asJSON bool) (string, error)
 }
 
+// Judge is implemented by clients whose provider classifies natively. Its
+// confidence is a calibrated probability rather than a number the model
+// reports about itself, and the inputs travel apart from the question, so
+// matchers use it over Complete when it is available.
+type Judge interface {
+	Judge(ctx context.Context, question string, labels, inputs []string) (Verdict, error)
+}
+
+// Verdict is the label a Judge chose. Model is the exact model version that
+// answered, when the provider reports one.
+type Verdict struct {
+	Label      string
+	Confidence float64
+	Model      string
+}
+
 // TruncateApproxTokens trims input to roughly maxTokens using the common
 // 4-chars-per-token approximation. Exact counting is provider-specific; this
 // only needs to keep a large body from blowing the context window.

pkg/operators/extractors/compile_test.go

@@ -130,3 +130,12 @@ func TestCompileExtractorsIgnoresXPathOnOtherTypes(t *testing.T) {
 	}
 	require.NoError(t, e.CompileExtractors())
 }
+
+func TestCompileExtractorsXPathUndeclaredVariable(t *testing.T) {
+	// Regression test: this expression used to compile and then panic
+	// with a nil pointer dereference inside antchfx/xpath when the
+	// first response was evaluated. It must fail at compile time.
+	var operator Extractor
+	require.NoError(t, yaml.Unmarshal([]byte("type: xpath\nxpath:\n  - \"******$A\"\n"), &operator))
+	require.ErrorContains(t, operator.CompileExtractors(), "undeclared variable")
+}

pkg/operators/matchers/compile_test.go

@@ -53,3 +53,12 @@ func TestCompileMatchersAllowsEmptyStringValue(t *testing.T) {
 	matched, _ := m.MatchWords("body", nil)
 	require.True(t, matched)
 }
+
+func TestCompileMatchersDSLTrailingBackslash(t *testing.T) {
+	// Regression test: a DSL expression whose string literal ends in a
+	// backslash used to panic the govaluate lexer (index out of range)
+	// while a template was loading. It must surface as a compile error.
+	var operator Matcher
+	require.NoError(t, yaml.Unmarshal([]byte("type: dsl\ndsl:\n  - \"body == 'x\\\\'\"\n"), &operator))
+	require.ErrorContains(t, operator.CompileMatchers(), "Unclosed string literal")
+}

pkg/operators/matchers/match_llm.go

@@ -65,26 +65,21 @@ func (matcher *Matcher) MatchLLMWithAudit(corpus string, inputs []string, values
 		return false, nil, nil
 	}
 
-	prompt := matcher.buildLLMPrompt(matcher.truncate(corpus, inputs), values)
-	answer, err := client.Complete(context.Background(), prompt, true)
-	if err != nil {
+	responses := matcher.truncate(corpus, inputs)
+	var (
+		verdict llmVerdict
+		audit   *LLMAudit
+		ok      bool
+	)
+	if judge, isJudge := client.(llmclient.Judge); isJudge {
+		verdict, audit, ok = matcher.judgeLLM(judge, responses, values)
+	} else {
+		verdict, audit, ok = matcher.completeLLM(client, responses, values)
+	}
+	if !ok {
 		return false, nil, nil
 	}
 
-	var verdict llmVerdict
-	if err := json.Unmarshal([]byte(answer), &verdict); err != nil {
-		return false, nil, nil
-	}
-
-	audit := &LLMAudit{
-		PromptHash: hashPrompt(prompt),
-		Verdict:    strings.TrimSpace(verdict.Verdict),
-		Confidence: verdict.Confidence,
-	}
-	if namer, ok := client.(modelNamer); ok {
-		audit.Model = namer.Model()
-	}
-
 	expect := matcher.Expect
 	if expect == "" {
 		expect = "yes"
@@ -107,9 +102,55 @@ func (matcher *Matcher) MatchLLMWithAudit(corpus string, inputs []string, values
 		return false, nil, audit
 	}
 
+	if verdict.Evidence == "" {
+		return true, nil, audit
+	}
 	return true, []string{verdict.Evidence}, audit
 }
 
+// completeLLM asks a chat model for the verdict as JSON.
+func (matcher *Matcher) completeLLM(client LLMClient, inputs []string, values map[string]interface{}) (llmVerdict, *LLMAudit, bool) {
+	prompt := matcher.buildLLMPrompt(inputs, values)
+	answer, err := client.Complete(context.Background(), prompt, true)
+	if err != nil {
+		return llmVerdict{}, nil, false
+	}
+
+	var verdict llmVerdict
+	if err := json.Unmarshal([]byte(answer), &verdict); err != nil {
+		return llmVerdict{}, nil, false
+	}
+
+	audit := &LLMAudit{
+		PromptHash: hashPrompt(prompt),
+		Verdict:    strings.TrimSpace(verdict.Verdict),
+		Confidence: verdict.Confidence,
+	}
+	if namer, ok := client.(modelNamer); ok {
+		audit.Model = namer.Model()
+	}
+	return verdict, audit, true
+}
+
+// judgeLLM asks a native classifier. It needs no prompt framing: the question
+// and the responses are separate fields of the request, and the answer is a
+// probability over the allowed verdicts, not text. It quotes no evidence.
+func (matcher *Matcher) judgeLLM(judge llmclient.Judge, inputs []string, values map[string]interface{}) (llmVerdict, *LLMAudit, bool) {
+	question := interpolate(matcher.Prompt, values)
+	answer, err := judge.Judge(context.Background(), question, matcher.Options, inputs)
+	if err != nil {
+		return llmVerdict{}, nil, false
+	}
+
+	audit := &LLMAudit{
+		Model:      answer.Model,
+		PromptHash: hashPrompt(strings.Join(append([]string{question, strings.Join(matcher.Options, ",")}, inputs...), "\x00")),
+		Verdict:    answer.Label,
+		Confidence: answer.Confidence,
+	}
+	return llmVerdict{Verdict: answer.Label, Confidence: answer.Confidence}, audit, true
+}
+
 // buildLLMPrompt wraps the author's question with the output contract and the
 // response under a random boundary (see FrameResponse).
 //

pkg/operators/matchers/match_llm_test.go

@@ -6,6 +6,7 @@ import (
 	"regexp"
 	"testing"
 
+	llmclient "github.com/projectdiscovery/nuclei/v3/pkg/operators/common/llm"
 	"github.com/stretchr/testify/require"
 )
 
@@ -277,3 +278,61 @@ func TestMatchLLMWithoutInputsUsesPart(t *testing.T) {
 	require.Contains(t, stub.lastGot, "the body")
 	require.NotContains(t, stub.lastGot, "Response 1")
 }
+
+type stubJudge struct {
+	stubLLM
+	verdict  llmclient.Verdict
+	err      error
+	question string
+	labels   []string
+	inputs   []string
+}
+
+func (s *stubJudge) Judge(_ context.Context, question string, labels, inputs []string) (llmclient.Verdict, error) {
+	s.question, s.labels, s.inputs = question, labels, inputs
+	return s.verdict, s.err
+}
+
+func TestMatchLLMPrefersJudgeAndKeepsResponseOutOfQuestion(t *testing.T) {
+	m := llmMatcher("login page for {{Hostname}}?")
+	judge := &stubJudge{verdict: llmclient.Verdict{Label: "yes", Confidence: 0.97, Model: "jev-1.14.0"}}
+	m.SetLLMClient(judge)
+
+	ok, snips, audit := m.MatchLLMWithAudit("<form>", nil, map[string]interface{}{"Hostname": "example.com"})
+	require.True(t, ok)
+	require.Empty(t, snips, "a classifier quotes no evidence")
+	require.Zero(t, judge.calls, "Complete must not be used when the client can judge")
+	require.Equal(t, "login page for example.com?", judge.question)
+	require.Nil(t, judge.labels, "no options means a yes or no question")
+	require.Equal(t, []string{"<form>"}, judge.inputs)
+	require.Equal(t, "jev-1.14.0", audit.Model)
+	require.InDelta(t, 0.97, audit.Confidence, 1e-9)
+}
+
+func TestMatchLLMJudgeAppliesExpectOptionsAndConfidence(t *testing.T) {
+	m := llmMatcher("same access?")
+	m.Options = []string{"same", "different"}
+	m.Expect = "different"
+	m.MinConfidence = 0.9
+	judge := &stubJudge{verdict: llmclient.Verdict{Label: "different", Confidence: 0.8}}
+	m.SetLLMClient(judge)
+
+	ok, _, audit := m.MatchLLMWithAudit("", []string{"a", "b"}, nil)
+	require.False(t, ok, "below min-confidence")
+	require.NotNil(t, audit)
+	require.Equal(t, []string{"same", "different"}, judge.labels)
+	require.Equal(t, []string{"a", "b"}, judge.inputs)
+
+	judge.verdict.Confidence = 0.95
+	ok, _, _ = m.MatchLLMWithAudit("", []string{"a", "b"}, nil)
+	require.True(t, ok)
+}
+
+func TestMatchLLMJudgeFailsClosed(t *testing.T) {
+	m := llmMatcher("q")
+	m.SetLLMClient(&stubJudge{err: errors.New("boom")})
+
+	ok, _, audit := m.MatchLLMWithAudit("body", nil, nil)
+	require.False(t, ok)
+	require.Nil(t, audit)
+}

pkg/output/output.go

@@ -87,11 +87,13 @@ type StandardWriter struct {
 	DisableStdout         bool
 	AddNewLinesOutputFile bool // by default this is only done for stdout
 	KeysToRedact          []string
+	// NoRedact skips the default credential keys; KeysToRedact still apply.
+	NoRedact bool
 
-	// redactPatterns caches the compiled credential-redaction regexes so they
+	// redactSet caches the compiled credential-redaction regexes so they
 	// are not recompiled on every Write call.
 	redactPatternsOnce sync.Once
-	redactPatterns     []*regexp.Regexp
+	redactSet          compiledRedaction
 
 	// JSONLogRequestHook is a hook that can be used to log request/response
 	// when using custom server code with output
@@ -125,6 +127,10 @@ type InternalWrappedEvent struct {
 	// Only applicable if interactsh is used
 	// This is used to avoid duplicate successful interactsh events
 	InteractshMatched atomic.Bool
+	// CatchAll reports whether the response is the page the host serves for
+	// any missing path. It probes the host on first call, so it is only asked
+	// once a template matched. Nil when calibration is off or does not apply.
+	CatchAll func() bool
 }
 
 func (iwe *InternalWrappedEvent) CloneShallow() *InternalWrappedEvent {
@@ -133,6 +139,7 @@ func (iwe *InternalWrappedEvent) CloneShallow() *InternalWrappedEvent {
 		Results:         nil,
 		OperatorsResult: nil,
 		UsesInteractsh:  iwe.UsesInteractsh,
+		CatchAll:        iwe.CatchAll,
 	}
 }
 
@@ -237,6 +244,25 @@ type ResultEvent struct {
 	FileToIndexPosition map[string]int `json:"-"`
 	TemplateVerifier    string         `json:"-"`
 	Error               string         `json:"error,omitempty"`
+	// TemplateInfo is the info as the template wrote it, kept only when
+	// placeholders changed the text. The changed fields can carry response data,
+	// so they are redacted like the raw request and response. It is bookkeeping,
+	// so no exporter writes it; mongo serializes with bson tags, not json ones.
+	TemplateInfo *model.Info `json:"-" bson:"-"`
+}
+
+// MarshalJSON encodes the event with every map nested in Metadata keyed by
+// strings. Metadata can hold javascript args decoded from template YAML.
+func (r ResultEvent) MarshalJSON() ([]byte, error) {
+	// resultEvent has the fields of ResultEvent but not this method, so marshaling it does not recurse.
+	type resultEvent ResultEvent
+	encoded := resultEvent(r)
+	metadata, err := types.ToJSONMap(r.Metadata)
+	if err != nil {
+		return nil, err
+	}
+	encoded.Metadata = metadata
+	return json.Marshal(encoded)
 }
 
 type IssueTrackerMetadata struct {
@@ -301,6 +327,7 @@ func NewStandardWriter(options *types.Options) (*StandardWriter, error) {
 		storeResponseDir:  options.StoreResponseDir,
 		omitTemplate:      options.OmitTemplate,
 		KeysToRedact:      options.Redact,
+		NoRedact:          options.NoRedact,
 	}
 
 	if v := os.Getenv("DISABLE_STDOUT"); v == "true" || v == "1" {
@@ -360,10 +387,13 @@ func (w *StandardWriter) Write(event *ResultEvent) error {
 	// user-supplied -redact keys. patterns are compiled once and cached to
 	// avoid recompiling regexes on every event in the write hot path.
 	patterns := w.redactionPatterns()
-	event.Request = redactWithPatterns(event.Request, patterns)
-	event.Response = redactWithPatterns(event.Response, patterns)
-	event.CURLCommand = redactWithPatterns(event.CURLCommand, patterns)
-	event.Matched = redactWithPatterns(event.Matched, patterns)
+	event.Request = redactWithSet(event.Request, patterns)
+	event.Response = redactWithSet(event.Response, patterns)
+	event.CURLCommand = redactWithSet(event.CURLCommand, patterns)
+	event.Matched = redactWithSet(event.Matched, patterns)
+	for _, field := range event.Info.FilledFields(event.TemplateInfo) {
+		*field.Value = redactWithSet(*field.Value, patterns)
+	}
 
 	event.Timestamp = time.Now()
 
@@ -404,8 +434,8 @@ func (w *StandardWriter) Write(event *ResultEvent) error {
 	return nil
 }
 
-// defaultRedactKeys are header and parameter names that are always redacted
-// from output regardless of the -redact flag.
+// defaultRedactKeys are header and parameter names that are redacted from
+// output regardless of the -redact flag, unless -no-redact is set.
 var defaultRedactKeys = []string{
 	"Authorization",
 	"Proxy-Authorization",
@@ -414,38 +444,94 @@ var defaultRedactKeys = []string{
 	"X-Api-Key",
 }
 
-// defaultRedactPatterns are the precompiled regexes for defaultRedactKeys.
-var defaultRedactPatterns = compileRedactPatterns(defaultRedactKeys)
+// cookieHeaderValue is the value of one Cookie or Set-Cookie header, stopping
+// at a quote so a curl -H 'Cookie: ...' does not swallow the rest of the line.
+var cookieHeaderValue = regexp.MustCompile(`(?i)((?:Cookie|Set-Cookie)\s*:\s*)([^"'\r\n]*)`)
+
+// compiledRedaction holds the three shapes a key can take. A header value is
+// masked to the end of the line. Inside a cookie header, key=value stops at
+// the next cookie delimiter so neighbouring cookies stay. A query value keeps
+// an embedded ';' and runs to '&', a quote, or whitespace.
+type compiledRedaction struct {
+	headers []*regexp.Regexp
+	cookies []*regexp.Regexp
+	queries []*regexp.Regexp
+}
+
+func (c compiledRedaction) empty() bool {
+	return len(c.headers) == 0 && len(c.cookies) == 0 && len(c.queries) == 0
+}
 
 // compileRedactPatterns builds the credential-redaction regexes for the given keys.
-func compileRedactPatterns(keys []string) []*regexp.Regexp {
-	patterns := make([]*regexp.Regexp, 0, len(keys))
+func compileRedactPatterns(keys []string) compiledRedaction {
+	var set compiledRedaction
 	for _, key := range keys {
-		patterns = append(patterns, regexp.MustCompile(fmt.Sprintf(`(?i)(%s\s*[:=]\s*["']?)[^"'\r\n&]+(["'\r\n]?)`, regexp.QuoteMeta(key))))
+		quoted := regexp.QuoteMeta(key)
+		set.headers = append(set.headers, regexp.MustCompile(fmt.Sprintf(`(?i)(%s\s*:\s*["']?)([^"'&\r\n]+)(["'\r\n]|$)`, quoted)))
+		set.cookies = append(set.cookies, regexp.MustCompile(fmt.Sprintf(`(?i)(%s\s*=\s*["']?)([^"'&\r\n\s]*?)(;\s*[\w.-]+(?:=|;|\r|\n|$)|["'&\r\n\s]|$)`, quoted)))
+		set.queries = append(set.queries, regexp.MustCompile(fmt.Sprintf(`(?i)(%s\s*=\s*["']?)([^"'&\r\n\s]*)(["'&\r\n\s]|$)`, quoted)))
 	}
-	return patterns
+	return set
 }
 
-// redactionPatterns returns the cached set of redaction patterns for this
-// writer (default keys plus any user-supplied -redact keys), compiling them on
-// first use.
-func (w *StandardWriter) redactionPatterns() []*regexp.Regexp {
+// redactionPatterns returns the cached redaction set for this writer (default
+// keys unless -no-redact is set, plus any user-supplied -redact keys).
+func (w *StandardWriter) redactionPatterns() compiledRedaction {
 	w.redactPatternsOnce.Do(func() {
-		if len(w.KeysToRedact) == 0 {
-			w.redactPatterns = defaultRedactPatterns
-			return
+		keys := make([]string, 0, len(defaultRedactKeys)+len(w.KeysToRedact))
+		if !w.NoRedact {
+			keys = append(keys, defaultRedactKeys...)
 		}
-		patterns := make([]*regexp.Regexp, 0, len(defaultRedactPatterns)+len(w.KeysToRedact))
-		patterns = append(patterns, defaultRedactPatterns...)
-		patterns = append(patterns, compileRedactPatterns(w.KeysToRedact)...)
-		w.redactPatterns = patterns
+		keys = append(keys, w.KeysToRedact...)
+		w.redactSet = compileRedactPatterns(keys)
 	})
-	return w.redactPatterns
+	return w.redactSet
 }
 
-func redactWithPatterns(data string, patterns []*regexp.Regexp) string {
+func redactKeys(data string, keys []string) string {
+	return redactWithSet(data, compileRedactPatterns(keys))
+}
+
+func redactWithSet(data string, set compiledRedaction) string {
+	if data == "" || set.empty() {
+		return data
+	}
+	data = applyRedact(data, set.headers)
+	if len(set.cookies) > 0 {
+		data = cookieHeaderValue.ReplaceAllStringFunc(data, func(match string) string {
+			parts := cookieHeaderValue.FindStringSubmatch(match)
+			if len(parts) != 3 {
+				return match
+			}
+			return parts[1] + applyRedact(parts[2], set.cookies)
+		})
+	}
+	return redactOutsideCookieValues(data, set.queries)
+}
+
+func redactOutsideCookieValues(data string, patterns []*regexp.Regexp) string {
+	if len(patterns) == 0 {
+		return data
+	}
+	locs := cookieHeaderValue.FindAllStringSubmatchIndex(data, -1)
+	if len(locs) == 0 {
+		return applyRedact(data, patterns)
+	}
+	var b strings.Builder
+	prev := 0
+	for _, loc := range locs {
+		valueStart, valueEnd := loc[4], loc[5]
+		b.WriteString(applyRedact(data[prev:valueStart], patterns))
+		b.WriteString(data[valueStart:valueEnd])
+		prev = valueEnd
+	}
+	b.WriteString(applyRedact(data[prev:], patterns))
+	return b.String()
+}
+
+func applyRedact(data string, patterns []*regexp.Regexp) string {
 	for _, pattern := range patterns {
-		data = pattern.ReplaceAllString(data, `$1***$2`)
+		data = pattern.ReplaceAllString(data, `$1***$3`)
 	}
 	return data
 }

pkg/output/output_test.go

@@ -6,8 +6,13 @@ import (
 	"testing"
 
 	"github.com/pkg/errors"
+	"github.com/projectdiscovery/goflags"
+	"github.com/projectdiscovery/nuclei/v3/pkg/model"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/projectdiscovery/nuclei/v3/pkg/utils/json"
+	"github.com/projectdiscovery/nuclei/v3/pkg/utils/yaml"
 	"github.com/stretchr/testify/require"
+	"go.mongodb.org/mongo-driver/bson"
 )
 
 func TestStandardWriterRequest(t *testing.T) {
@@ -51,6 +56,152 @@ func TestStandardWriterRequest(t *testing.T) {
 	})
 }
 
+// Only the fields filled from a response are redacted; text the template wrote
+// stays as it is, even next to a filled field.
+func TestStandardWriterRedactsFilledInfo(t *testing.T) {
+	w, err := NewStandardWriter(&types.Options{Redact: goflags.StringSlice{"password"}})
+	require.NoError(t, err)
+	w.DisableStdout = true
+
+	written := model.Info{Description: "login with password: {{password}}", Remediation: "Rotate the password: it is public"}
+	filled := &ResultEvent{TemplateInfo: &written, Info: model.Info{Description: "login with password: admin", Remediation: written.Remediation}}
+	static := &ResultEvent{Info: model.Info{Description: "send the password: header"}}
+	require.NoError(t, w.Write(filled))
+	require.NoError(t, w.Write(static))
+
+	require.Equal(t, "login with password: ***", filled.Info.Description)
+	require.Equal(t, "Rotate the password: it is public", filled.Info.Remediation)
+	require.Equal(t, "send the password: header", static.Info.Description)
+}
+
+// Credential headers are masked by default; -no-redact keeps them while keys
+// given with -redact are still masked.
+func TestStandardWriterNoRedact(t *testing.T) {
+	const request = "GET /?tracking=1&x=2 HTTP/1.1\r\nCookie: tracking=1; session=abc\r\nX-Token: secret\r\n\r\n"
+	const response = "HTTP/1.1 200 OK\r\nSet-Cookie: session=abc; Path=/\r\n\r\n"
+	const curl = "curl -H 'Cookie: tracking=1; session=abc' https://example.com/?tracking=1&x=2"
+	const matched = "https://example.com/?tracking=1&x=2"
+
+	cases := []struct {
+		name     string
+		options  *types.Options
+		request  string
+		response string
+		curl     string
+		matched  string
+	}{
+		{
+			"default", &types.Options{},
+			"GET /?tracking=1&x=2 HTTP/1.1\r\nCookie: ***\r\nX-Token: secret\r\n\r\n",
+			"HTTP/1.1 200 OK\r\nSet-Cookie: ***\r\n\r\n",
+			"curl -H 'Cookie: ***' https://example.com/?tracking=1&x=2",
+			matched,
+		},
+		{"no-redact", &types.Options{NoRedact: true}, request, response, curl, matched},
+		{
+			"no-redact with header key", &types.Options{NoRedact: true, Redact: goflags.StringSlice{"X-Token"}},
+			"GET /?tracking=1&x=2 HTTP/1.1\r\nCookie: tracking=1; session=abc\r\nX-Token: ***\r\n\r\n",
+			response, curl, matched,
+		},
+		// One cookie or query parameter is masked; its neighbours stay.
+		{
+			"no-redact with cookie key", &types.Options{NoRedact: true, Redact: goflags.StringSlice{"tracking"}},
+			"GET /?tracking=***&x=2 HTTP/1.1\r\nCookie: tracking=***; session=abc\r\nX-Token: secret\r\n\r\n",
+			response,
+			"curl -H 'Cookie: tracking=***; session=abc' https://example.com/?tracking=***&x=2",
+			"https://example.com/?tracking=***&x=2",
+		},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			w, err := NewStandardWriter(tc.options)
+			require.NoError(t, err)
+			w.DisableStdout = true
+
+			event := &ResultEvent{Request: request, Response: response, CURLCommand: curl, Matched: matched}
+			require.NoError(t, w.Write(event))
+
+			require.Equal(t, tc.request, event.Request)
+			require.Equal(t, tc.response, event.Response)
+			require.Equal(t, tc.curl, event.CURLCommand)
+			require.Equal(t, tc.matched, event.Matched)
+		})
+	}
+}
+
+// A `;` inside a query value is part of the value; in a cookie header it is a delimiter.
+func TestRedactKeepsSemicolonInsideValue(t *testing.T) {
+	cases := map[string]string{
+		"GET /?password=;s3cr3t&x=2 HTTP/1.1":             "GET /?password=***&x=2 HTTP/1.1",
+		"GET /?password=a;b HTTP/1.1":                     "GET /?password=*** HTTP/1.1",
+		"GET /?password=super;secret=hidden&x=1 HTTP/1.1": "GET /?password=***&x=1 HTTP/1.1",
+		"https://example.com/?password=a;b":               "https://example.com/?password=***",
+		"https://example.com/?password=a;session=abc":     "https://example.com/?password=***",
+		"Cookie: password=x; session=abc\r\n":             "Cookie: password=***; session=abc\r\n",
+		"Set-Cookie: password=x; Path=/; HttpOnly\r\n":    "Set-Cookie: password=***; Path=/; HttpOnly\r\n",
+		"Set-Cookie: password=x; HttpOnly\r\n":            "Set-Cookie: password=***; HttpOnly\r\n",
+		"password=a;b&user=u":                             "password=***&user=u",
+		"curl -H 'Cookie: password=x; session=abc' 'https://example.com/?password=super;secret=hidden&x=1'": "curl -H 'Cookie: password=***; session=abc' 'https://example.com/?password=***&x=1'",
+	}
+	for in, want := range cases {
+		require.Equal(t, want, redactKeys(in, []string{"password"}), in)
+	}
+}
+
+// The mongo exporter serializes results with bson, which ignores json tags.
+func TestResultEventKeepsTemplateInfoOutOfBSON(t *testing.T) {
+	written := model.Info{Description: "{{host}} is exposed"}
+	data, err := bson.Marshal(&ResultEvent{TemplateInfo: &written, Info: model.Info{Description: "example.com is exposed"}})
+	require.NoError(t, err)
+
+	var doc bson.M
+	require.NoError(t, bson.Unmarshal(data, &doc))
+	require.NotContains(t, doc, "templateinfo")
+	require.Contains(t, doc, "info")
+}
+
+// Template YAML decodes nested maps with interface keys, in the info metadata and
+// in the javascript args a result carries as its metadata.
+func TestFormatJSONNestedYAMLMaps(t *testing.T) {
+	var template struct {
+		Info model.Info             `yaml:"info"`
+		Args map[string]interface{} `yaml:"args"`
+	}
+	err := yaml.Unmarshal([]byte(`
+info:
+  name: Nested metadata
+  author: repro
+  severity: info
+  metadata:
+    labels:
+      en: Hello
+    items:
+      - fr: Bonjour
+args:
+  Nested:
+    en: Hello
+`), &template)
+	require.NoError(t, err)
+
+	w, err := NewStandardWriter(&types.Options{JSONL: true})
+	require.NoError(t, err)
+	data, err := w.formatJSON(&ResultEvent{TemplateID: "nested-metadata", Info: template.Info, Metadata: template.Args})
+	require.NoError(t, err)
+
+	var decoded struct {
+		Info struct {
+			Metadata map[string]interface{} `json:"metadata"`
+		} `json:"info"`
+		Meta map[string]interface{} `json:"meta"`
+	}
+	require.NoError(t, json.Unmarshal(data, &decoded))
+	require.Equal(t, map[string]interface{}{
+		"labels": map[string]interface{}{"en": "Hello"},
+		"items":  []interface{}{map[string]interface{}{"fr": "Bonjour"}},
+	}, decoded.Info.Metadata)
+	require.Equal(t, map[string]interface{}{"Nested": map[string]interface{}{"en": "Hello"}}, decoded.Meta)
+}
+
 type testWriteCloser struct {
 	strings.Builder
 }

pkg/protocols/common/expressions/expressions_test.go

@@ -38,6 +38,9 @@ func TestEvaluate(t *testing.T) {
 		{input: "{{base64(rawhash)}}", expected: "{{contact_id}}{{email}}", extra: map[string]any{
 			"rawhash": `{"contact_id":"{{contact_id}}","email":"{{email}}"}`,
 		}},
+		// regression: string concatenation of variables inside base64() must not
+		// stringify the parameter list (govaluate produced "[admin:]admin" here)
+		{input: "{{base64(username + ':' + password)}}", expected: "YWRtaW46YWRtaW4=", extra: map[string]interface{}{"username": "admin", "password": "admin"}},
 	}
 	for _, item := range items {
 		value, err := Evaluate(item.input, item.extra)

pkg/protocols/common/protocolstate/exclude_race_test.go

@@ -0,0 +1,46 @@
+package protocolstate
+
+import (
+	"sync"
+	"testing"
+
+	"github.com/projectdiscovery/networkpolicy"
+	"github.com/stretchr/testify/require"
+)
+
+func TestIsHostAllowedRacesExcludeUpdate(t *testing.T) {
+	np, err := networkpolicy.New(networkpolicy.Options{DenyList: []string{"127.0.0.0/8"}})
+	require.NoError(t, err)
+
+	const executionID = "exclude-race"
+	entry := &Dialers{NetworkPolicy: np, ExcludeTargets: []string{"192.0.2.1"}}
+	err = dialers.Set(executionID, entry)
+	require.NoError(t, err)
+	t.Cleanup(func() { dialers.Delete(executionID) })
+
+	stop := make(chan struct{})
+	var wg sync.WaitGroup
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		lists := [][]string{{"192.0.2.1"}, {"192.0.2.2", "192.0.2.3"}}
+		i := 0
+		for {
+			select {
+			case <-stop:
+				return
+			default:
+			}
+			entry.Lock()
+			entry.ExcludeTargets = lists[i%2]
+			entry.Unlock()
+			i++
+		}
+	}()
+
+	for i := 0; i < 2000; i++ {
+		_ = IsHostAllowed(executionID, "http://192.0.2.1")
+	}
+	close(stop)
+	wg.Wait()
+}

pkg/protocols/common/protocolstate/headless.go

@@ -102,33 +102,28 @@ func isValidHost(options *types.Options, targetUrl string) bool {
 	if !stringsutil.HasPrefixAny(normalized, "http:", "https:", "ws:", "wss:") {
 		return true
 	}
-
-	dialers, ok := dialers.Get(options.ExecutionId)
-	if !ok {
+	if options == nil {
 		return true
 	}
-
-	np := dialers.NetworkPolicy
-	if np == nil {
-		// no policy present: deny rather than allow.
-		return false
-	}
-
-	urlx, err := urlutil.Parse(targetUrl)
-	if err != nil {
-		// not a valid url
-		return false
-	}
-	targetUrl = urlx.Hostname()
-	_, ok = np.ValidateHost(targetUrl)
-	return ok
+	return IsHostAllowed(options.ExecutionId, targetUrl)
 }
 
 // CanonicalizeHost normalizes a hostname or IP string:
 // - strips surrounding whitespace and brackets (e.g. "[::1]" -> "::1")
-// - maps IPv4-mapped IPv6 literals (e.g. "::ffff:127.0.0.1" or "0:0:0:0:0:ffff:127.0.0.1") to canonical IPv4 ("127.0.0.1")
+// - maps IPv4-mapped IPv6 literals (e.g. "::ffff:127.0.0.1" or "[::ffff:127.0.0.1]:80") to canonical IPv4 ("127.0.0.1" or "127.0.0.1:80")
 func CanonicalizeHost(host string) string {
 	h := strings.TrimSpace(host)
+	if sh, port, err := net.SplitHostPort(h); err == nil {
+		sh = strings.TrimPrefix(sh, "[")
+		sh = strings.TrimSuffix(sh, "]")
+		if ip := net.ParseIP(sh); ip != nil {
+			if ipv4 := ip.To4(); ipv4 != nil {
+				return net.JoinHostPort(ipv4.String(), port)
+			}
+			return net.JoinHostPort(ip.String(), port)
+		}
+		return h
+	}
 	h = strings.TrimPrefix(h, "[")
 	h = strings.TrimSuffix(h, "]")
 	if ip := net.ParseIP(h); ip != nil {
@@ -143,11 +138,14 @@ func CanonicalizeHost(host string) string {
 // IsHostAllowed checks if the host is allowed by network policy
 func IsHostAllowed(executionId string, targetUrl string) bool {
 	dialers, ok := dialers.Get(executionId)
-	if !ok {
+	if !ok || dialers == nil {
 		return true
 	}
 
+	dialers.Lock()
 	np := dialers.NetworkPolicy
+	excludedTargets := append([]string(nil), dialers.ExcludeTargets...)
+	dialers.Unlock()
 	if np == nil {
 		// no policy present: deny rather than allow.
 		return false
@@ -179,7 +177,7 @@ func IsHostAllowed(executionId string, targetUrl string) bool {
 	canonHost := CanonicalizeHost(host)
 
 	// Check explicit ExcludeTargets against raw, host, and canonical host
-	for _, excluded := range dialers.ExcludeTargets {
+	for _, excluded := range excludedTargets {
 		ex := CanonicalizeHost(strings.TrimSpace(excluded))
 		if ex != "" {
 			if strings.EqualFold(canonHost, ex) || strings.EqualFold(host, ex) || strings.EqualFold(rawTarget, ex) {
@@ -188,15 +186,11 @@ func IsHostAllowed(executionId string, targetUrl string) bool {
 		}
 	}
 
-	// Validate host against NetworkPolicy using both canonical and original host
+	// The policy stores IPv4 prefixes. A mapped form such as ::ffff:203.0.113.10
+	// has to be judged as 203.0.113.10, or an allow-listed address is rejected.
 	if _, ok := np.ValidateHost(canonHost); !ok {
 		return false
 	}
-	if canonHost != host {
-		if _, ok := np.ValidateHost(host); !ok {
-			return false
-		}
-	}
 	return true
 }
 
@@ -221,14 +215,35 @@ func SetupHeadlessNetworkPolicy(options *types.Options, page *rod.Page) error {
 			target = parsed.Hostname()
 		}
 		if !strings.Contains(target, "/") {
-			blockedPatterns = append(blockedPatterns,
-				"*://"+target+"/*",
-				"*://"+target+":*/*",
-				"ws://"+target+"/*",
-				"ws://"+target+":*/*",
-				"wss://"+target+"/*",
-				"wss://"+target+":*/*",
-			)
+			addPatterns := func(t string) {
+				t = strings.TrimPrefix(strings.TrimSuffix(t, "]"), "[")
+				blockedPatterns = append(blockedPatterns,
+					"*://"+t+"/*",
+					"*://"+t+":*/*",
+					"ws://"+t+"/*",
+					"ws://"+t+":*/*",
+					"wss://"+t+"/*",
+					"wss://"+t+":*/*",
+				)
+				if ip := net.ParseIP(t); ip != nil {
+					if ipv4 := ip.To4(); ipv4 != nil {
+						v6mapped := "[::ffff:" + ipv4.String() + "]"
+						blockedPatterns = append(blockedPatterns,
+							"*://"+v6mapped+"/*",
+							"*://"+v6mapped+":*/*",
+							"ws://"+v6mapped+"/*",
+							"ws://"+v6mapped+":*/*",
+							"wss://"+v6mapped+"/*",
+							"wss://"+v6mapped+":*/*",
+						)
+					}
+				}
+			}
+			addPatterns(target)
+			canon := CanonicalizeHost(target)
+			if canon != "" && canon != target {
+				addPatterns(canon)
+			}
 		}
 	}
 
@@ -255,13 +270,32 @@ func SetupHeadlessNetworkPolicy(options *types.Options, page *rod.Page) error {
 			"wss://10.*/*", "wss://10.*:*/*",
 			"wss://192.168.*/*", "wss://192.168.*:*/*",
 			"wss://169.254.*/*", "wss://169.254.*:*/*",
+			"*://[::ffff:127.*]/*", "*://[::ffff:127.*]:*/*",
+			"*://[::ffff:0.0.0.0]/*", "*://[::ffff:0.0.0.0]:*/*",
+			"*://[::ffff:10.*]/*", "*://[::ffff:10.*]:*/*",
+			"*://[::ffff:192.168.*]/*", "*://[::ffff:192.168.*]:*/*",
+			"*://[::ffff:169.254.*]/*", "*://[::ffff:169.254.*]:*/*",
+			"ws://[::ffff:127.*]/*", "ws://[::ffff:127.*]:*/*",
+			"ws://[::ffff:0.0.0.0]/*", "ws://[::ffff:0.0.0.0]:*/*",
+			"ws://[::ffff:10.*]/*", "ws://[::ffff:10.*]:*/*",
+			"ws://[::ffff:192.168.*]/*", "ws://[::ffff:192.168.*]:*/*",
+			"ws://[::ffff:169.254.*]/*", "ws://[::ffff:169.254.*]:*/*",
+			"wss://[::ffff:127.*]/*", "wss://[::ffff:127.*]:*/*",
+			"wss://[::ffff:0.0.0.0]/*", "wss://[::ffff:0.0.0.0]:*/*",
+			"wss://[::ffff:10.*]/*", "wss://[::ffff:10.*]:*/*",
+			"wss://[::ffff:192.168.*]/*", "wss://[::ffff:192.168.*]:*/*",
+			"wss://[::ffff:169.254.*]/*", "wss://[::ffff:169.254.*]:*/*",
 		}
 		for i := 16; i <= 31; i++ {
 			prefix := fmt.Sprintf("172.%d.*", i)
+			v6prefix := fmt.Sprintf("[::ffff:172.%d.*]", i)
 			localPatterns = append(localPatterns,
 				"*://"+prefix+"/*", "*://"+prefix+":*/*",
+				"*://"+v6prefix+"/*", "*://"+v6prefix+":*/*",
 				"ws://"+prefix+"/*", "ws://"+prefix+":*/*",
+				"ws://"+v6prefix+"/*", "ws://"+v6prefix+":*/*",
 				"wss://"+prefix+"/*", "wss://"+prefix+":*/*",
+				"wss://"+v6prefix+"/*", "wss://"+v6prefix+":*/*",
 			)
 		}
 		blockedPatterns = append(blockedPatterns, localPatterns...)

pkg/protocols/common/protocolstate/headless_test.go

@@ -19,7 +19,8 @@ func TestIsValidHost(t *testing.T) {
 	require.NoError(t, err)
 
 	d := &Dialers{
-		NetworkPolicy: np,
+		NetworkPolicy:  np,
+		ExcludeTargets: []string{"192.168.1.50"},
 	}
 	err = dialers.Set(opts.ExecutionId, d)
 	require.NoError(t, err)
@@ -31,18 +32,75 @@ func TestIsValidHost(t *testing.T) {
 	require.True(t, isValidHost(opts, "ws://example.com:9999"))
 	require.True(t, isValidHost(opts, "wss://example.com:9999"))
 
-	// Blocked hosts (127.0.0.1)
+	// Blocked hosts (127.0.0.1 in DenyList)
 	require.False(t, isValidHost(opts, "http://127.0.0.1:8080"))
 	require.False(t, isValidHost(opts, "https://127.0.0.1:8443"))
 	require.False(t, isValidHost(opts, "ws://127.0.0.1:9999"))
 	require.False(t, isValidHost(opts, "wss://127.0.0.1:9999"))
 
+	// Blocked IPv4-mapped IPv6 forms of 127.0.0.1
+	require.False(t, isValidHost(opts, "http://[::ffff:127.0.0.1]:8080"))
+	require.False(t, isValidHost(opts, "https://[::ffff:127.0.0.1]:8443"))
+	require.False(t, isValidHost(opts, "http://[::ffff:7f00:1]:8080"))
+	require.False(t, isValidHost(opts, "http://[0:0:0:0:0:ffff:127.0.0.1]:8080"))
+	require.False(t, isValidHost(opts, "http://[0000:0000:0000:0000:0000:ffff:7f00:0001]:8080"))
+	require.False(t, isValidHost(opts, "ws://[::ffff:127.0.0.1]:9999"))
+
+	// Blocked hosts via ExcludeTargets (IPv4 and IPv4-mapped IPv6)
+	require.False(t, isValidHost(opts, "http://192.168.1.50:8080"))
+	require.False(t, isValidHost(opts, "http://[::ffff:192.168.1.50]:8080"))
+	require.False(t, isValidHost(opts, "http://[::ffff:c0a8:132]:8080"))
+
 	require.True(t, IsHostAllowed(opts.ExecutionId, "ws://example.com:9999"))
 	require.False(t, IsHostAllowed(opts.ExecutionId, "ws://127.0.0.1:9999"))
+	require.False(t, IsHostAllowed(opts.ExecutionId, "ws://[::ffff:127.0.0.1]:9999"))
+	require.False(t, IsHostAllowed(opts.ExecutionId, "http://[::ffff:192.168.1.50]:8080"))
+}
+
+func TestIsHostAllowedMappedFormOfAllowListedIP(t *testing.T) {
+	const executionID = "test-allow-mapped"
+	opts := &types.Options{ExecutionId: executionID}
+
+	np, err := networkpolicy.New(networkpolicy.Options{
+		AllowList: []string{"203.0.113.10"},
+		DenyList:  []string{"127.0.0.1"},
+	})
+	require.NoError(t, err)
+
+	err = dialers.Set(executionID, &Dialers{NetworkPolicy: np})
+	require.NoError(t, err)
+	defer dialers.Delete(executionID)
+
+	require.True(t, isValidHost(opts, "http://203.0.113.10"))
+	require.True(t, isValidHost(opts, "http://[::ffff:203.0.113.10]"))
+	require.True(t, isValidHost(opts, "http://[::ffff:cb00:710a]:8080"))
+	require.False(t, isValidHost(opts, "http://[::ffff:127.0.0.1]"))
+	require.False(t, isValidHost(opts, "http://198.51.100.1"))
+}
+
+func TestCanonicalizeHost_IPv4MappedIPv6(t *testing.T) {
+	cases := map[string]string{
+		"127.0.0.1":                                "127.0.0.1",
+		"127.0.0.1:8080":                           "127.0.0.1:8080",
+		"::ffff:127.0.0.1":                         "127.0.0.1",
+		"[::ffff:127.0.0.1]":                       "127.0.0.1",
+		"[::ffff:127.0.0.1]:8080":                  "127.0.0.1:8080",
+		"::ffff:7f00:1":                            "127.0.0.1",
+		"[::ffff:7f00:1]:8443":                     "127.0.0.1:8443",
+		"0:0:0:0:0:ffff:127.0.0.1":                 "127.0.0.1",
+		"0000:0000:0000:0000:0000:ffff:7f00:0001": "127.0.0.1",
+		"example.com":                              "example.com",
+		"example.com:443":                          "example.com:443",
+		"::1":                                      "::1",
+		"[::1]:8080":                               "[::1]:8080",
+	}
+
+	for input, expected := range cases {
+		require.Equal(t, expected, CanonicalizeHost(input), "input: %s", input)
+	}
 }
 
 func TestSetupHeadlessNetworkPolicy_Nil(t *testing.T) {
 	require.NoError(t, SetupHeadlessNetworkPolicy(nil, nil))
 	require.NoError(t, SetupHeadlessNetworkPolicy(&types.Options{}, nil))
 }
-

pkg/protocols/common/protocolstate/state.go

@@ -5,11 +5,14 @@ import (
 	"fmt"
 	"net"
 	"net/url"
+	"regexp"
 	"slices"
+	"strings"
 	"time"
 
 	"github.com/go-sql-driver/mysql"
 	"github.com/pkg/errors"
+	"golang.org/x/net/idna"
 	"golang.org/x/net/proxy"
 
 	"github.com/projectdiscovery/fastdialer/fastdialer"
@@ -18,6 +21,7 @@ import (
 	"github.com/projectdiscovery/nuclei/v3/pkg/protocols/http/httpcache"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
 	"github.com/projectdiscovery/nuclei/v3/pkg/utils/expand"
+	iputil "github.com/projectdiscovery/utils/ip"
 	mapsutil "github.com/projectdiscovery/utils/maps"
 )
 
@@ -101,15 +105,20 @@ func initDialers(options *types.Options) error {
 		switch {
 		case asn.IsASN(excludeTarget):
 			expandedDenyList = append(expandedDenyList, expand.ASN(excludeTarget)...)
-		default:
+		case iputil.IsIP(excludeTarget) || iputil.IsCIDR(excludeTarget):
 			expandedDenyList = append(expandedDenyList, excludeTarget)
+		default:
+			expandedDenyList = append(expandedDenyList, excludeHostRules(excludeTarget)...)
 		}
 	}
 
 	if options.RestrictLocalNetworkAccess {
 		expandedDenyList = append(expandedDenyList, networkpolicy.DefaultIPv4DenylistRanges...)
 		expandedDenyList = append(expandedDenyList, networkpolicy.DefaultIPv6DenylistRanges...)
+		// the unspecified address connects to the local host
+		expandedDenyList = append(expandedDenyList, "::/128")
 	}
+	expandedDenyList = append(expandedDenyList, mappedDenyRanges(expandedDenyList)...)
 	npOptions := &networkpolicy.Options{
 		DenyList: expandedDenyList,
 	}
@@ -335,3 +344,47 @@ func Close(executionId string) {
 		StopActiveMemGuardian()
 	}
 }
+
+// mappedDenyRanges returns the IPv4-mapped IPv6 form (::ffff:a.b.c.d) of each
+// IPv4 address and range in denyList. The network policy checks a mapped
+// address such as [::ffff:127.0.0.1] only against IPv6 ranges, but the dialer
+// connects to the IPv4 host, so without these forms it passes -eh and -lna.
+func mappedDenyRanges(denyList []string) []string {
+	var mapped []string
+	for _, entry := range denyList {
+		cidr := entry
+		if !strings.Contains(cidr, "/") {
+			cidr += "/32"
+		}
+		_, ipNet, err := net.ParseCIDR(cidr)
+		if err != nil || ipNet.IP.To4() == nil || strings.Contains(entry, ":") {
+			continue
+		}
+		ones, _ := ipNet.Mask.Size()
+		mapped = append(mapped, fmt.Sprintf("::ffff:%s/%d", ipNet.IP.To4(), 96+ones))
+	}
+	return mapped
+}
+
+// excludeHostRules turns an exclude entry into network policy deny rules.
+// networkpolicy treats non-IP entries as unanchored regexes, so a plain
+// "test.com" would also deny "latest.com". Plain hostnames match the host and
+// its subdomains instead; entries with regex syntax are kept as given.
+func excludeHostRules(target string) []string {
+	if strings.ContainsAny(target, `\^$*+?()[]{}|`) {
+		return []string{target}
+	}
+
+	host := strings.TrimSuffix(target, ".")
+	hosts := []string{host}
+	// the network DSL helpers only normalize plain rules, so add the IDNA form here
+	if ascii, err := idna.ToASCII(strings.ToLower(host)); err == nil && !strings.EqualFold(ascii, host) {
+		hosts = append(hosts, ascii)
+	}
+
+	rules := make([]string, 0, len(hosts))
+	for _, h := range hosts {
+		rules = append(rules, `(?i)(^|\.)`+regexp.QuoteMeta(h)+`\.?$`)
+	}
+	return rules
+}

pkg/protocols/common/protocolstate/state_test.go

@@ -0,0 +1,78 @@
+package protocolstate
+
+import (
+	"context"
+	"net"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+	"time"
+
+	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/stretchr/testify/require"
+)
+
+func TestExcludeTargetsMatchExactHost(t *testing.T) {
+	opts := types.DefaultOptions()
+	opts.ExecutionId = "test-exclude-targets"
+	opts.ExcludeTargets = []string{"test.com", "192.0.2.0/24", `^regex\.example$`, "bücher.example"}
+	require.NoError(t, Init(opts))
+	defer Close(opts.ExecutionId)
+
+	np := GetDialersWithId(opts.ExecutionId).NetworkPolicy
+	require.False(t, np.Validate("test.com"))
+	require.False(t, np.Validate("TEST.com"))
+	require.False(t, np.Validate("test.com."))
+	require.False(t, np.Validate("sub.test.com"))
+	require.False(t, np.Validate("192.0.2.10"))
+	require.False(t, np.Validate("regex.example"))
+	require.False(t, np.Validate("xn--bcher-kva.example"))
+	require.True(t, np.Validate("latest.com"))
+	require.True(t, np.Validate("test.com.example.org"))
+	require.True(t, np.Validate("testXcom"))
+	require.True(t, np.Validate("sub.regex.example"))
+}
+
+func TestMappedFormsDeniedByDialer(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {}))
+	defer server.Close()
+	_, port, err := net.SplitHostPort(server.Listener.Addr().String())
+	require.NoError(t, err)
+
+	for name, set := range map[string]func(*types.Options){
+		"lna":           func(o *types.Options) { o.RestrictLocalNetworkAccess = true },
+		"exclude-hosts": func(o *types.Options) { o.ExcludeTargets = []string{"127.0.0.1"} },
+		"exclude-cidr":  func(o *types.Options) { o.ExcludeTargets = []string{"127.0.0.0/8"} },
+	} {
+		t.Run(name, func(t *testing.T) {
+			opts := types.DefaultOptions()
+			opts.ExecutionId = "test-mapped-" + name
+			set(opts)
+			require.NoError(t, Init(opts))
+			defer Close(opts.ExecutionId)
+
+			d := GetDialersWithId(opts.ExecutionId)
+			hosts := []string{"127.0.0.1", "::ffff:127.0.0.1", "::ffff:7f00:1", "0:0:0:0:0:ffff:127.0.0.1"}
+			if opts.RestrictLocalNetworkAccess {
+				hosts = append(hosts, "::")
+			}
+			for _, host := range hosts {
+				require.False(t, d.NetworkPolicy.Validate(host), host)
+				ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+				conn, err := d.Fastdialer.Dial(ctx, "tcp", net.JoinHostPort(host, port))
+				cancel()
+				if conn != nil {
+					_ = conn.Close()
+				}
+				require.Error(t, err, host)
+			}
+			require.True(t, d.NetworkPolicy.Validate("::ffff:8.8.8.8"))
+			require.True(t, d.NetworkPolicy.Validate("8.8.8.8"))
+		})
+	}
+}
+
+func TestMappedDenyRanges(t *testing.T) {
+	require.Equal(t, []string{"::ffff:127.0.0.0/104", "::ffff:192.0.2.10/128", "::ffff:0.0.0.0/104"},
+		mappedDenyRanges([]string{"127.0.0.0/8", "192.0.2.10", "::1/128", "fc00::/7", `(?i)(^|\.)test\.com\.?$`, "0.0.0.0/8"}))
+}

pkg/protocols/dns/dnsclientpool/clientpool.go

@@ -26,33 +26,42 @@ var defaultResolvers = []string{
 	"8.8.4.4:53", // Google
 }
 
-// Init initializes the client pool implementation
+// Init initializes the client pool implementation.
+// Each resolver list gets its own client, so a later scan is not stuck with the first scan's resolvers.
 func Init(options *types.Options) error {
-	m.Lock()
-	defer m.Unlock()
-
-	// Don't create clients if already created in the past.
-	if normalClient != nil {
-		return nil
-	}
-	clientPool = mapsutil.NewSyncLockMap[string, *retryabledns.Client]()
-
-	resolvers := defaultResolvers
-	if len(options.InternalResolversList) > 0 {
-		resolvers = options.InternalResolversList
-	}
-	var err error
-	normalClient, err = retryabledns.New(resolvers, 1)
+	client, err := cachedClient(resolversFor(options))
 	if err != nil {
 		return errors.Wrap(err, "could not create dns client")
 	}
+	m.Lock()
+	normalClient = client
+	m.Unlock()
 	return nil
 }
 
-func getNormalClient() *retryabledns.Client {
+func resolversFor(options *types.Options) []string {
+	if options != nil && len(options.InternalResolversList) > 0 {
+		return options.InternalResolversList
+	}
+	return defaultResolvers
+}
+
+func cachedClient(resolvers []string) (*retryabledns.Client, error) {
 	m.Lock()
 	defer m.Unlock()
-	return normalClient
+	if clientPool == nil {
+		clientPool = mapsutil.NewSyncLockMap[string, *retryabledns.Client]()
+	}
+	key := strings.Join(resolvers, ",")
+	if client, ok := clientPool.Get(key); ok {
+		return client, nil
+	}
+	client, err := retryabledns.New(resolvers, 1)
+	if err != nil {
+		return nil, err
+	}
+	_ = clientPool.Set(key, client)
+	return client, nil
 }
 
 // Configuration contains the custom configuration options for a client
@@ -81,7 +90,7 @@ func (c *Configuration) Hash() string {
 // Get creates or gets a client for the protocol based on custom configuration
 func Get(options *types.Options, configuration *Configuration) (*retryabledns.Client, error) {
 	if (configuration.Retries <= 1) && len(configuration.Resolvers) == 0 {
-		return getNormalClient(), nil
+		return cachedClient(resolversFor(options))
 	}
 	hash := configuration.Hash()
 	if client, ok := clientPool.Get(hash); ok {

pkg/protocols/dns/dnsclientpool/clientpool_test.go

@@ -0,0 +1,35 @@
+package dnsclientpool
+
+import (
+	"testing"
+
+	"github.com/projectdiscovery/nuclei/v3/pkg/types"
+	"github.com/stretchr/testify/require"
+)
+
+func TestClientsFollowResolverList(t *testing.T) {
+	previousClient := normalClient
+	previousPool := clientPool
+	t.Cleanup(func() {
+		normalClient = previousClient
+		clientPool = previousPool
+	})
+	normalClient = nil
+	clientPool = nil
+
+	firstOptions := &types.Options{InternalResolversList: []string{"192.0.2.1:53"}}
+	secondOptions := &types.Options{InternalResolversList: []string{"192.0.2.2:53"}}
+
+	require.NoError(t, Init(firstOptions))
+	require.NoError(t, Init(secondOptions))
+
+	first, err := Get(firstOptions, &Configuration{})
+	require.NoError(t, err)
+	second, err := Get(secondOptions, &Configuration{})
+	require.NoError(t, err)
+	require.NotSame(t, first, second)
+
+	again, err := Get(firstOptions, &Configuration{})
+	require.NoError(t, err)
+	require.Same(t, first, again)
+}

pkg/protocols/headless/engine/engine.go

@@ -12,6 +12,7 @@ import (
 	"github.com/go-rod/rod/lib/launcher/flags"
 	"github.com/pkg/errors"
 
+	"github.com/projectdiscovery/gologger"
 	"github.com/projectdiscovery/nuclei/v3/pkg/types"
 	fileutil "github.com/projectdiscovery/utils/file"
 	osutils "github.com/projectdiscovery/utils/os"
@@ -38,7 +39,7 @@ func New(options *types.Options) (*Browser, error) {
 	var proxyServer *headlessProxy
 	var err error
 
-	chromeLauncher := launcher.New()
+	var chromeLauncher *launcher.Launcher
 
 	if options.CDPEndpoint == "" {
 		dataStore, err = os.MkdirTemp("", "nuclei-*")
@@ -46,58 +47,33 @@ func New(options *types.Options) (*Browser, error) {
 			return nil, errors.Wrap(err, "could not create temporary directory")
 		}
 
-		chromeLauncher = chromeLauncher.
-			Leakless(false).
-			Set("disable-crash-reporter").
-			Set("disable-gpu").
-			Set("disable-notifications").
-			Set("hide-scrollbars").
-			Set("ignore-certificate-errors").
-			Set("ignore-ssl-errors").
-			Set("incognito").
-			Set("mute-audio").
-			Set("window-size", fmt.Sprintf("%d,%d", 1080, 1920)).
-			Delete("use-mock-keychain").
-			UserDataDir(dataStore)
-
-		if MustDisableSandbox() {
-			chromeLauncher = chromeLauncher.NoSandbox(true)
-		}
-
-		executablePath, err := os.Executable()
-		if err != nil {
-			return nil, err
-		}
-
-		// if musl is used, most likely we are on alpine linux which is not supported by go-rod, so we fallback to default chrome
-		useMusl, _ := fileutil.UseMusl(executablePath)
-		if options.UseInstalledChrome || useMusl {
-			if chromePath, hasChrome := launcher.LookPath(); hasChrome {
-				chromeLauncher.Bin(chromePath)
-			} else {
-				return nil, errors.New("the chrome browser is not installed")
-			}
-		}
-
-		if options.ShowBrowser {
-			chromeLauncher = chromeLauncher.Headless(false)
-		} else {
-			chromeLauncher = chromeLauncher.Headless(true)
-		}
 		proxyServer, err = newHeadlessProxy(options)
 		if err != nil {
 			return nil, err
 		}
-		chromeLauncher = chromeLauncher.
-			Proxy(proxyServer.Addr()).
-			Set("proxy-bypass-list", "<-loopback>")
 
-		for k, v := range options.ParseHeadlessOptionalArguments() {
-			chromeLauncher.Set(flags.Flag(k), v)
+		noSandbox := MustDisableSandbox()
+		chromeLauncher, err = newChromeLauncher(options, dataStore, proxyServer.Addr(), noSandbox)
+		if err != nil {
+			_ = proxyServer.Close()
+			return nil, err
 		}
 
 		launcherURL, err = chromeLauncher.Launch()
+		if err != nil && !noSandbox && !chromeLauncher.Has(flags.NoSandbox) && isSandboxLaunchError(err) {
+			logger := options.Logger
+			if logger == nil {
+				logger = gologger.DefaultLogger
+			}
+			logger.Warning().Msgf("Could not start the browser with its sandbox, starting it without sandbox. To keep the sandbox, allow unprivileged user namespaces (for example sysctl kernel.unprivileged_userns_clone=1 or kernel.apparmor_restrict_unprivileged_userns=0) or in docker use a seccomp profile that allows user namespaces (for example --security-opt seccomp=unconfined): %s", sandboxLine(err.Error()))
+			chromeLauncher.Kill()
+			chromeLauncher, err = newChromeLauncher(options, dataStore, proxyServer.Addr(), true)
+			if err == nil {
+				launcherURL, err = chromeLauncher.Launch()
+			}
+		}
 		if err != nil {
+			_ = proxyServer.Close()
 			return nil, err
 		}
 	} else {
@@ -143,11 +119,79 @@ func New(options *types.Options) (*Browser, error) {
 	return engine, nil
 }
 
-// MustDisableSandbox determines if the current os and user needs sandbox mode disabled
+// MustDisableSandbox reports whether the browser must run without its sandbox.
+// Chrome refuses to start sandboxed as root on Linux, so only that case needs
+// "--no-sandbox". Other users keep the sandbox; New falls back to no sandbox
+// with a warning when the system can not provide one.
+// https://github.com/chromium/chromium/blob/c4d3c31083a2e1481253ff2d24298a1dfe19c754/chrome/test/chromedriver/client/chromedriver.py#L209
 func MustDisableSandbox() bool {
-	// linux with root user needs "--no-sandbox" option
-	// https://github.com/chromium/chromium/blob/c4d3c31083a2e1481253ff2d24298a1dfe19c754/chrome/test/chromedriver/client/chromedriver.py#L209
-	return osutils.IsLinux()
+	return mustDisableSandbox(osutils.IsLinux(), os.Geteuid())
+}
+
+func mustDisableSandbox(linux bool, euid int) bool {
+	return linux && euid == 0
+}
+
+// newChromeLauncher returns the launcher for a local browser. User headless
+// options are applied last, so "-headless-options no-sandbox" still works.
+func newChromeLauncher(options *types.Options, dataStore, proxyAddr string, noSandbox bool) (*launcher.Launcher, error) {
+	chromeLauncher := launcher.New().
+		Leakless(false).
+		Set("disable-crash-reporter").
+		Set("disable-gpu").
+		Set("disable-notifications").
+		Set("hide-scrollbars").
+		Set("ignore-certificate-errors").
+		Set("ignore-ssl-errors").
+		Set("incognito").
+		Set("mute-audio").
+		Set("window-size", fmt.Sprintf("%d,%d", 1080, 1920)).
+		Delete("use-mock-keychain").
+		UserDataDir(dataStore).
+		// rod adds "--no-sandbox" by default inside containers, so the
+		// sandbox is always set explicitly.
+		NoSandbox(noSandbox)
+
+	executablePath, err := os.Executable()
+	if err != nil {
+		return nil, err
+	}
+
+	// if musl is used, most likely we are on alpine linux which is not supported by go-rod, so we fallback to default chrome
+	useMusl, _ := fileutil.UseMusl(executablePath)
+	if options.UseInstalledChrome || useMusl {
+		if chromePath, hasChrome := launcher.LookPath(); hasChrome {
+			chromeLauncher.Bin(chromePath)
+		} else {
+			return nil, errors.New("the chrome browser is not installed")
+		}
+	}
+
+	chromeLauncher = chromeLauncher.
+		Headless(!options.ShowBrowser).
+		Proxy(proxyAddr).
+		Set("proxy-bypass-list", "<-loopback>")
+
+	for k, v := range options.ParseHeadlessOptionalArguments() {
+		chromeLauncher.Set(flags.Flag(k), v)
+	}
+	return chromeLauncher, nil
+}
+
+// isSandboxLaunchError reports whether the browser exited because its
+// sandbox could not be set up, for example "No usable sandbox!".
+func isSandboxLaunchError(err error) bool {
+	return err != nil && strings.Contains(strings.ToLower(err.Error()), "sandbox")
+}
+
+// sandboxLine returns the browser output line about the sandbox.
+func sandboxLine(value string) string {
+	for _, line := range strings.Split(value, "\n") {
+		if strings.Contains(strings.ToLower(line), "sandbox") {
+			return strings.TrimSpace(line)
+		}
+	}
+	return strings.TrimSpace(value)
 }
 
 // SetUserAgent sets custom user agent to the browser

pkg/protocols/headless/engine/headers.go

@@ -0,0 +1,82 @@
+package engine
+
+import (
+	"maps"
+	"net/http"
+	"slices"
+	"strings"
+)
+
+// reservedResponseKeys are the event fields that the headless protocol owns.
+// A response header or cookie that normalizes to one of them is dropped, so a
+// target cannot spoof them (the HTTP protocol also lets its own fields win).
+var reservedResponseKeys = map[string]struct{}{
+	"template-id":   {},
+	"template-info": {},
+	"template-path": {},
+	"host":          {},
+	"matched":       {},
+	"type":          {},
+	"req":           {},
+	"resp":          {},
+	"body":          {},
+	"data":          {},
+	"header":        {},
+	"all_headers":   {},
+	"status_code":   {},
+	"history":       {},
+	"duration":      {},
+	// read by the result event, so a target must not set them
+	"ip":      {},
+	"path":    {},
+	"request": {},
+	"error":   {},
+}
+
+// addResponseVariables exposes response headers and cookies as DSL variables
+// (same normalization as the HTTP protocol).
+// Header names are lowercased with "-" replaced by "_", e.g. Content-Type -> content_type.
+// Cookie names are lowercased as-is.
+// browserCookies are the cookies the browser holds for the page, as the
+// browser does not report Set-Cookie in intercepted response headers. A
+// Set-Cookie of the response wins over a browser cookie, and a header wins
+// over a cookie of the same name. Headers that normalize to the same name are
+// applied in sorted order, so the result is deterministic. Keys already in
+// data and reserved protocol keys are never overwritten.
+func addResponseVariables(data map[string]interface{}, resp *http.Response, browserCookies []*http.Cookie) {
+	if resp == nil {
+		return
+	}
+	vars := make(map[string]string)
+	for _, cookie := range browserCookies {
+		vars[strings.ToLower(cookie.Name)] = cookie.Value
+	}
+	for _, cookie := range resp.Cookies() {
+		vars[strings.ToLower(cookie.Name)] = cookie.Value
+	}
+	for _, name := range slices.Sorted(maps.Keys(resp.Header)) {
+		k := strings.ToLower(strings.ReplaceAll(strings.TrimSpace(name), "-", "_"))
+		vars[k] = strings.Join(resp.Header[name], " ")
+	}
+	for k, v := range vars {
+		if _, ok := reservedResponseKeys[k]; ok || strings.HasPrefix(k, "duration_") {
+			continue
+		}
+		if _, ok := data[k]; ok {
+			continue
+		}
+		data[k] = v
+	}
+}
+
+// cookieMatchesHost reports whether a browser cookie with the given domain is
+// sent to host. A host-only cookie has a domain without a leading dot and only
+// matches that exact host; a domain cookie also matches its subdomains.
+func cookieMatchesHost(domain, host string) bool {
+	host = strings.ToLower(host)
+	domain = strings.ToLower(domain)
+	if trimmed, ok := strings.CutPrefix(domain, "."); ok {
+		return host == trimmed || strings.HasSuffix(host, domain)
+	}
+	return host == domain
+}

pkg/protocols/headless/engine/headers_test.go

@@ -0,0 +1,97 @@
+package engine
+
+import (
+	"net/http"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestAddResponseVariables(t *testing.T) {
+	resp := &http.Response{Header: make(http.Header)}
+	resp.Header.Set("Content-Type", "text/plain; charset=utf-8")
+	resp.Header.Add("X-Custom", "a")
+	resp.Header.Add("X-Custom", "b")
+	resp.Header.Add("Set-Cookie", "SessionID=abc; Path=/")
+	resp.Header.Add("Set-Cookie", "theme=dark; Path=/")
+
+	data := make(map[string]interface{})
+	addResponseVariables(data, resp, nil)
+
+	require.Equal(t, "text/plain; charset=utf-8", data["content_type"])
+	require.Equal(t, "a b", data["x_custom"])
+	require.Equal(t, "abc", data["sessionid"])
+	require.Equal(t, "dark", data["theme"])
+	require.NotContains(t, data, "header")
+}
+
+func TestAddResponseVariablesNil(t *testing.T) {
+	data := make(map[string]interface{})
+	addResponseVariables(data, nil, nil)
+	require.Empty(t, data)
+}
+
+func TestAddResponseVariablesCollisions(t *testing.T) {
+	resp := &http.Response{Header: make(http.Header)}
+	resp.Header.Set("Status-Code", "999")
+	resp.Header.Set("Data", "spoofed")
+	resp.Header.Set("Host", "evil.example")
+	resp.Header.Set("Duration-1", "0")
+	resp.Header.Set("IP", "6.6.6.6")
+	resp.Header.Set("Path", "/spoofed")
+	resp.Header.Set("Request", "spoofed")
+	resp.Header.Set("Error", "spoofed")
+	resp.Header.Set("X-Action", "from header")
+	resp.Header.Add("Set-Cookie", "header=spoofed; Path=/")
+	resp.Header.Add("Set-Cookie", "x_token=cookie; Path=/")
+	resp.Header.Set("X-Token", "header")
+	// both names normalize to x_dup; sorted order makes the result stable
+	resp.Header["X-Dup"] = []string{"dash"}
+	resp.Header["X_dup"] = []string{"underscore"}
+
+	data := map[string]interface{}{
+		"header":      "real headers",
+		"status_code": "200",
+		"x_action":    "from action",
+	}
+	addResponseVariables(data, resp, nil)
+
+	require.Equal(t, "real headers", data["header"])
+	require.Equal(t, "200", data["status_code"])
+	require.Equal(t, "from action", data["x_action"])
+	require.NotContains(t, data, "data")
+	require.NotContains(t, data, "host")
+	require.NotContains(t, data, "duration_1")
+	for _, key := range []string{"ip", "path", "request", "error"} {
+		require.NotContains(t, data, key)
+	}
+	require.Equal(t, "header", data["x_token"])
+	require.Equal(t, "underscore", data["x_dup"])
+}
+
+func TestAddResponseVariablesBrowserCookies(t *testing.T) {
+	resp := &http.Response{Header: make(http.Header)}
+	resp.Header.Add("Set-Cookie", "theme=dark; Path=/")
+	browserCookies := []*http.Cookie{
+		{Name: "SessID", Value: "abc"},
+		{Name: "theme", Value: "light"},
+		{Name: "data", Value: "spoofed"},
+	}
+
+	data := make(map[string]interface{})
+	addResponseVariables(data, resp, browserCookies)
+
+	require.Equal(t, "abc", data["sessid"])
+	require.Equal(t, "dark", data["theme"])
+	require.NotContains(t, data, "data")
+}
+
+func TestCookieMatchesHost(t *testing.T) {
+	require.True(t, cookieMatchesHost("example.com", "example.com"))
+	require.False(t, cookieMatchesHost("example.com", "sub.example.com"))
+	require.True(t, cookieMatchesHost(".example.com", "example.com"))
+	require.True(t, cookieMatchesHost(".example.com", "sub.example.com"))
+	require.False(t, cookieMatchesHost(".example.com", "badexample.com"))
+	require.False(t, cookieMatchesHost("other.com", "example.com"))
+	require.True(t, cookieMatchesHost("127.0.0.1", "127.0.0.1"))
+}

pkg/protocols/headless/engine/page.go

@@ -212,28 +212,39 @@ func (i *Instance) Run(ctx *contextargs.Context, actions []*Action, payloads map
 		return nil, nil, err
 	}
 
-	if !options.DisableCookie {
-		// at the end of actions pull out updated cookies from the browser and inject them into the shared cookie jar
-		if cookies, err := page.Cookies([]string{URL.String()}); !options.DisableCookie && err == nil && len(cookies) > 0 {
-			var httpCookies []*http.Cookie
-			for _, cookie := range cookies {
-				httpCookie := &http.Cookie{
-					Name:     cookie.Name,
-					Value:    cookie.Value,
-					Domain:   cookie.Domain,
-					Path:     cookie.Path,
-					HttpOnly: cookie.HTTPOnly,
-					Secure:   cookie.Secure,
-				}
-				httpCookies = append(httpCookies, httpCookie)
+	// at the end of actions pull out the cookies from the browser. They are
+	// exposed as DSL variables, since the browser does not report Set-Cookie
+	// in the intercepted response headers, and they are injected into the
+	// shared cookie jar unless cookie reuse is disabled. The cookies are read
+	// for the whole host, because a cookie scoped to another path than the
+	// input URL is still set by the response.
+	var httpCookies []*http.Cookie
+	if cookies, err := page.Browser().GetCookies(); err == nil {
+		for _, cookie := range cookies {
+			if !cookieMatchesHost(cookie.Domain, URL.Hostname()) {
+				continue
 			}
-			ctx.CookieJar.SetCookies(URL, httpCookies)
+			httpCookie := &http.Cookie{
+				Name:     cookie.Name,
+				Value:    cookie.Value,
+				Domain:   cookie.Domain,
+				Path:     cookie.Path,
+				HttpOnly: cookie.HTTPOnly,
+				Secure:   cookie.Secure,
+			}
+			httpCookies = append(httpCookies, httpCookie)
 		}
 	}
+	if !options.DisableCookie && len(httpCookies) > 0 {
+		ctx.CookieJar.SetCookies(URL, httpCookies)
+	}
 
 	if resp := createdPage.FirstResponse(); resp != nil {
-		data["header"] = utils.HeadersToString(resp.Header)
+		headers := utils.HeadersToString(resp.Header)
+		data["header"] = headers
+		data["all_headers"] = headers
 		data["status_code"] = fmt.Sprint(resp.StatusCode)
+		addResponseVariables(data, resp, httpCookies)
 	}
 
 	successfulPageCreation = true // avoid closing the page in case of success in deferred function

pkg/protocols/headless/engine/page_actions_test.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"io"
 	"math/rand"
+	"net"
 	"net/http"
 	"net/http/cookiejar"
 	"net/http/httptest"
@@ -15,6 +16,7 @@ import (
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"testing"
 	"time"
 
@@ -319,6 +321,58 @@ func TestActionHeadersChange(t *testing.T) {
 	})
 }
 
+// Chrome keeps malformed percent-escapes such as "100%.png" in request paths,
+// as the WHATWG URL Standard requires, but net/url rejects them, so go-rod
+// hands the modification-rules hijack handler a request with a nil URL (#8015).
+//
+//nolint:exhaustruct,exhaustruct_v5,paralleltest // like the other action tests, which share browser and dialer state
+func TestRoutingRuleHandlerFailsRequestNetHTTPCannotParse(t *testing.T) {
+	for _, disableCookie := range []bool{false, true} {
+		t.Run(fmt.Sprintf("DisableCookie=%t", disableCookie), func(t *testing.T) {
+			// Count connections, not requests: net/http rejects the request line
+			// before any handler runs. TLS makes the headless proxy tunnel the
+			// request with CONNECT instead of rejecting the unparsable path itself.
+			var imageConns atomic.Int32
+
+			imageServer := httptest.NewUnstartedServer(http.NotFoundHandler())
+			imageServer.Config.ConnState = func(_ net.Conn, state http.ConnState) {
+				if state == http.StateNew {
+					imageConns.Add(1)
+				}
+			}
+
+			imageServer.StartTLS()
+			defer imageServer.Close()
+
+			actions := []*Action{
+				{
+					ActionType: ActionTypeHolder{ActionType: ActionSetHeader},
+					Data:       map[string]string{"part": "request", "key": "Test", "value": "Hello"},
+				},
+				{ActionType: ActionTypeHolder{ActionType: ActionNavigate}, Data: map[string]string{"url": "{{BaseURL}}"}},
+				{ActionType: ActionTypeHolder{ActionType: ActionWaitLoad}},
+			}
+			handler := func(w http.ResponseWriter, r *http.Request) {
+				if r.Header.Get("Test") == "Hello" {
+					_, _ = fmt.Fprintf(w, `<html><body>found<img src="%s/100%%.png"></body></html>`, imageServer.URL)
+				}
+			}
+			options := &Options{Timeout: 20 * time.Second, DisableCookie: disableCookie}
+
+			testHeadlessWithOptions(t, actions, options, handler, func(page *Page, err error, _ ActionData) {
+				require.NoError(t, err)
+
+				body := page.Page().MustElement("body").MustText()
+				require.Equal(t, "found", strings.TrimSpace(body), "modification rules were not applied")
+				require.Zero(t, imageConns.Load(), "unparsable request reached the image server")
+
+				imageDone := page.Page().MustEval(`() => document.querySelector("img").complete`).Bool()
+				require.True(t, imageDone, "unparsable request is still paused")
+			})
+		})
+	}
+}
+
 func TestActionScreenshot(t *testing.T) {
 	response := `
 		<html>
@@ -1003,6 +1057,21 @@ func TestMain(m *testing.M) {
 func testHeadless(t *testing.T, actions []*Action, timeout time.Duration, handler func(w http.ResponseWriter, r *http.Request), assert func(page *Page, pageErr error, extractedData ActionData)) {
 	t.Helper()
 
+	options := &Options{Timeout: timeout} //nolint:exhaustruct,exhaustruct_v5 // zero values are the defaults
+	testHeadlessWithOptions(t, actions, options, handler, assert)
+}
+
+// testHeadlessWithOptions is testHeadless with caller-chosen page options. It
+// overwrites options.Options with the test scan options.
+func testHeadlessWithOptions(
+	t *testing.T,
+	actions []*Action,
+	options *Options,
+	handler func(w http.ResponseWriter, r *http.Request),
+	assert func(page *Page, pageErr error, extractedData ActionData),
+) {
+	t.Helper()
+
 	lfa := envutil.GetEnvOrDefault("LOCAL_FILE_ACCESS", true)
 	rna := envutil.GetEnvOrDefault("RESTRICTED_LOCAL_NETWORK_ACCESS", false)
 
@@ -1027,7 +1096,8 @@ func testHeadless(t *testing.T, actions []*Action, timeout time.Duration, handle
 	input.CookieJar, err = cookiejar.New(nil)
 	require.Nil(t, err)
 
-	extractedData, page, err := instance.Run(input, actions, nil, &Options{Timeout: timeout, Options: opts}) // allow file access in test
+	options.Options = opts // allow file access in test
+	extractedData, page, err := instance.Run(input, actions, nil, options)
 	assert(page, err, extractedData)
 
 	if page != nil {
@@ -1156,3 +1226,22 @@ func TestBlockedHeadlessURLS(t *testing.T) {
 		}
 	}
 }
+
+func TestBrowserCookiesOutsideInputPath(t *testing.T) {
+	actions := []*Action{
+		{ActionType: ActionTypeHolder{ActionType: ActionNavigate}, Data: map[string]string{"url": "{{BaseURL}}"}},
+		{ActionType: ActionTypeHolder{ActionType: ActionWaitLoad}},
+	}
+
+	handler := func(w http.ResponseWriter, r *http.Request) {
+		http.SetCookie(w, &http.Cookie{Name: "session", Value: "s1", Path: "/"})
+		http.SetCookie(w, &http.Cookie{Name: "token", Value: "abc", Path: "/admin"})
+		_, _ = fmt.Fprintln(w, `ok`)
+	}
+
+	testHeadless(t, actions, 20*time.Second, handler, func(page *Page, err error, out ActionData) {
+		require.Nil(t, err, "could not run page actions")
+		require.Equal(t, "s1", out["session"])
+		require.Equal(t, "abc", out["token"], "a cookie scoped to another path must still be exposed")
+	})
+}

pkg/protocols/headless/engine/proxy_test.go

@@ -56,7 +56,10 @@ func TestHeadlessProxy_BlocksExcludedHosts(t *testing.T) {
 				return
 			}
 			allowedHits.Add(1)
+			reqBuf := make([]byte, 512)
+			_, _ = conn.Read(reqBuf)
 			_, _ = conn.Write([]byte("HTTP/1.1 200 OK\r\nContent-Length: 2\r\n\r\nOK"))
+			time.Sleep(50 * time.Millisecond)
 			_ = conn.Close()
 		}
 	}()

pkg/protocols/headless/engine/rules.go

@@ -15,8 +15,21 @@ import (
 // routingRuleHandler handles proxy rule for actions related to request/response modification
 func (p *Page) routingRuleHandler(httpClient *http.Client) func(ctx *rod.Hijack) {
 	return func(ctx *rod.Hijack) {
+		req := ctx.Request.Req()
+		// go-rod discards url.Parse errors, so a URL that Chrome accepts but
+		// net/url rejects, such as "/100%.png", arrives as nil. Fail the request:
+		// the Go client cannot send it, and continuing it in the browser would
+		// drop the modification rules and, with -cdp-endpoint (no headless
+		// proxy), the network policy's dial-time address checks.
+		if req.URL == nil {
+			gologger.Verbose().Msgf("headless: failed %s request with a URL that net/url cannot parse", req.Method)
+			ctx.Response.Fail(proto.NetworkErrorReasonFailed)
+
+			return
+		}
+
 		// usually browsers don't use chunked transfer encoding, so we set the content-length nevertheless
-		ctx.Request.Req().ContentLength = int64(len(ctx.Request.Body()))
+		req.ContentLength = int64(len(ctx.Request.Body()))
 		// snapshot the rules once: ExecuteActions may still be appending rules
 		// from another goroutine while this hijack handler runs.
 		rules := p.rulesSnapshot()
@@ -28,17 +41,17 @@ func (p *Page) routingRuleHandler(httpClient *http.Client) func(ctx *rod.Hijack)
 			switch rule.Action {
 			case ActionSetMethod:
 				rule.Do(func() {
-					ctx.Request.Req().Method = rule.Args["method"]
+					req.Method = rule.Args["method"]
 				})
 			case ActionAddHeader:
-				ctx.Request.Req().Header.Add(rule.Args["key"], rule.Args["value"])
+				req.Header.Add(rule.Args["key"], rule.Args["value"])
 			case ActionSetHeader:
-				ctx.Request.Req().Header.Set(rule.Args["key"], rule.Args["value"])
+				req.Header.Set(rule.Args["key"], rule.Args["value"])
 			case ActionDeleteHeader:
-				ctx.Request.Req().Header.Del(rule.Args["key"])
+				req.Header.Del(rule.Args["key"])
 			case ActionSetBody:
 				body := rule.Args["body"]
-				ctx.Request.Req().ContentLength = int64(len(body))
+				req.ContentLength = int64(len(body))
 				ctx.Request.SetBody(body)
 			}
 		}
@@ -46,21 +59,21 @@ func (p *Page) routingRuleHandler(httpClient *http.Client) func(ctx *rod.Hijack)
 		// each http request is performed via the native go http client
 		// we first inject the shared cookies
 		if !p.options.DisableCookie {
-			if cookies := p.ctx.CookieJar.Cookies(ctx.Request.URL()); len(cookies) > 0 {
-				httpClient.Jar.SetCookies(ctx.Request.URL(), cookies)
+			if cookies := p.ctx.CookieJar.Cookies(req.URL); len(cookies) > 0 {
+				httpClient.Jar.SetCookies(req.URL, cookies)
 			}
 		}
 
 		// perform the request
 		if err := ctx.LoadResponse(httpClient, true); err != nil {
-			gologger.Verbose().Msgf("headless: failed to load response for %s: %s", ctx.Request.URL(), err)
+			gologger.Verbose().Msgf("headless: failed to load response for %s: %s", req.URL, err)
 		}
 
 		if !p.options.DisableCookie {
 			// retrieve the updated cookies from the native http client and inject them into the shared cookie jar
 			// keeps existing one if not present
-			if cookies := httpClient.Jar.Cookies(ctx.Request.URL()); len(cookies) > 0 {
-				p.ctx.CookieJar.SetCookies(ctx.Request.URL(), cookies)
+			if cookies := httpClient.Jar.Cookies(req.URL); len(cookies) > 0 {
+				p.ctx.CookieJar.SetCookies(req.URL, cookies)
 			}
 		}
 
@@ -84,7 +97,6 @@ func (p *Page) routingRuleHandler(httpClient *http.Client) func(ctx *rod.Hijack)
 		}
 
 		// store history
-		req := ctx.Request.Req()
 		var rawReq string
 		if raw, err := httputil.DumpRequestOut(req, true); err == nil {
 			rawReq = string(raw)