* test: assert absolute-path facts without frozen sentences
Sibling path tests now require the field path, invalid_format, and regex
expectation, and they record that each negative case ran. NUL rejection
follows the argument that contains the byte instead of searching for the
word NUL.
* test: retain actionable NUL path diagnostics
---------
Co-authored-by: Cursor Agent <[email protected]>
The help renderer appends "(default: X)" whenever the option field declares a
schema default, so descriptions that also spell the default out print it twice:
--block-size <value> Power-of-two block size from 256 to 65536 bytes
(default 4096) (default: 4096)
Remove the hand-written default from the eleven options that own a schema
default across inspect-byte-entropy, inspect-devicetree,
inspect-packet-capture, inspect-packet-fields, inspect-signal-decoders and
decode-signal-recording, matching the rule in cliDecimalField's contract that
the domain field owns bounds, defaults and help.
Options without a schema default keep their prose: --match_limit, --line_limit,
--limit and --packet_limit are nullable or optional, so the renderer emits no
annotation and the description is the only place the effective value appears.
Add a boundary test asserting each such option states its default exactly once,
so a restated default cannot come back unnoticed.
trace_call_path and trace_native_values both use workflow revision 2.
Value-trace absence stays specific to trace_native_values, and an
unrevised operation still matches the default profile.
Expose the existing opt-in Ghidra checkpoint lifecycle in tool descriptions and summary capabilities. Distinguish confirmed initial analysis from restored Programs, preserving the checkpoint path and digest through shared CLI/MCP results.
Remove prose-driven checks and their scaffolding; retain canonical effects and lifecycle/source-fact coverage. Validated with 445 focused tests, 13 real Ghidra CLI/MCP scenarios and three isolated local Codex CLI tasks covering disabled setup, saving edits and fresh-agent restoration.
* fix(cli): advertise domain constraints for selector and path options
Android class, method, and mapping fields, ADB setting keys, and browser
endpoints used bare strings in --schema. They now reuse the domain fields,
so empty selectors, relative mapping paths, and malformed setting keys
fail in the advertised schema.
* fix(cli): keep selector limits when argument parsing rejects them
Empty browser and Electron selectors now fail in the CLI parser, so their
diagnostics are VALIDATION_ERROR with the one-character minimum. ADB
setting keys stay on the selector repair path, which restores maximum 256
and the identifier pattern.
* test(cli): assert empty-selector codes without frozen prose
Empty browser and Electron selectors still fail as VALIDATION_ERROR
too_small on the named field. The tests no longer copy Zod's sentence.
* test(cli): compare endpoint diagnostics with the domain schema
Browser and Inspector tests take custom issue text from the schema that
owns it, and empty capture comparison must name both missing inspection
fields.
* fix(cli): remove duplicate merged schema import
---------
Co-authored-by: Cursor Agent <[email protected]>
* fix(mcp): advertise absolute paths for tools that reject relative inputs
The domain schemas accepted relative paths, then handlers rejected them
with isAbsolute. Those inputs now use absoluteLocalPathStringSchema, and
the duplicate handler checks are gone.
* fix(javascript): own recovery path admission in the schema
Relative recovery paths now fail in absoluteLocalPathStringSchema, including
the additional_paths index. Drop the duplicate handler check, and make the
CLI, MCP, and tools/list tests assert that schema fact.
* test: assert absolute-path facts instead of error prose
Recovery and web-script tests keep the field, format code, and regex
expectation. They no longer freeze the schema's explanatory sentence.
---------
Co-authored-by: Cursor Agent <[email protected]>
* fix(javascript): keep timer cancellation within Node's handle family
Node keeps timeouts and immediates in disjoint handle families. clearTimeout
and clearInterval are aliases that act only on a Timeout from setTimeout or
setInterval, while clearImmediate acts only on an Immediate from setImmediate.
Crossing the families returns without doing anything, so the scheduled
callback still runs.
REA linked a cancellation whenever the cleared binding traced back to any timer
schedule, so clearImmediate(setTimeout(work, 10)) reported a complete
cancellation and the graph emitted a resolved cancels-timer relation. It also
flagged no uncertainty for it.
Require the cancel and the schedule to share a handle family before linking
them. A cross-family clear now keeps linkedTimerId null and resolution
unresolved, and the graph reports the cancellation as a candidate with an
ambiguous-target unknown instead of a resolved fact.
Verified against Node v24.22.1 over all nine schedule/cancel pairs: setTimeout
and setInterval with clearTimeout or clearInterval cancel, and setImmediate
with clearImmediate cancels; the other five combinations leave the callback
running. Node's own lib/timers.js shows why: clearTimeout requires
timer?._onTimeout, clearImmediate requires immediate?._onImmediate and returns
early otherwise, and clearInterval delegates to clearTimeout.
* chore: retrigger title and readiness checks after the title was corrected
* fix(javascript): say cross-family timer clearing is runtime-dependent
Review feedback from @morluto: the previous wording claimed clearing across the
timer handle families "returns without doing anything, so the timer keeps
running". That does not hold on every supported Node version. Measured on Node
24.22.1, all three cross-family pairs fire:
| scheduled with | cancelled with | Node 24.22.1 |
| --- | --- | --- |
| `setTimeout` | `clearImmediate` | fires |
| `setImmediate` | `clearTimeout` | fires |
| `setImmediate` | `clearInterval` | fires |
`clearImmediate` cancelling a `Timeout` on Node 22 is the case the old claim
missed. Within a family the clear always cancels, which is why the analysis
links only same-family pairs; across families the outcome depends on the
runtime, which is exactly why the analysis reports `unresolved` rather than
`complete` in either direction.
The behavior and the tests are unchanged. The `timerHandleFamily` doc comment,
the linked-origin comment, and the test comment now state the runtime
dependence and why it justifies an unresolved result.
Add opt-in REA_GHIDRA_CHECKPOINT_DIR for owned Ghidra Program persistence through the shared CLI/MCP provider. Save completed initial analysis and metadata changes on normal close; verify and restore the last successful packed Program without auto-analysis.
Preserve generation identity, exclusive ownership, typed cancellation and cleanup retry. Refresh checkpoint-dependent profiles after retirement, including failed-switch restoration, while retaining caller selection provenance. Keep source executables and caller projects unchanged.
Validated with 720 focused tests (3 platform skips), 13 real Ghidra CLI/MCP scenarios, static/docs checks and CI. Review findings addressed. No version bump or release publication.
Refs #8165.
* feat(desktop): read XML with a plain DOCTYPE, and check an element's whole text
- Both readers accept a DOCTYPE with only a name and an optional PUBLIC
or SYSTEM identifier, record it as the view's doctype, and never fetch
the DTD or expand entities beyond XML's five; an internal subset, an
external entity and a parameter entity stay refused alike.
- text_equals and text_contains compare the concatenated text of the
element (or text node or attribute) a pointer reaches, as XPath
string() does, with optional normalize_whitespace; both evaluators
agree, and a text check on a truncated element is undecided.
- Whitespace-only text between elements is dropped only when the element
holds no other text, so text split into runs keeps its spaces.
- The draft suggests text_equals for an element whose text changed.
So an AbiWord .abw reads as XML, and its paragraph's text is checked
whole however AbiWord splits it into <c> runs, as Writer's spans are.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* fix(desktop): draft whole-text checks on the paragraph, read only text as text, and skip asctime stamps
- A changed value is read at both ends, so text that became a run still
marks its paragraph, and the drafted text_equals moves up out of mixed
content to the nearest ancestor named by an attribute, such as
AbiWord's p by its xid.
- Text checks accept only a text node or an attribute value; element
names and doctype fields fail, in both evaluators.
- isTimestamp also recognises asctime dates (Www Mmm dd hh:mm:ss yyyy).
- XML files are drafted with the keyed diff the archive entries use, so
no suggestion reaches an element by position; the shared logic moves to
appTaskDraftNamed.ts.
- Text-check tests assert facts, not whole sentences.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* fix(desktop): keep checks on a changed attribute inside a paragraph whose text check replaces its runs'
A drafted whole-text check on a paragraph hid every change inside it, so a
link whose label and href both changed drafted only the paragraph's text and
the new destination went unchecked. The text check now covers text and runs
added or removed whole; a changed name or attribute of an element both
versions hold keeps its own check.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* fix(desktop): read an attribute called name or attributes as an attribute when the draft keeps its check
The owner of `.../attributes/name` was taken as `.../attributes`, so a
changed `name` attribute inside a paragraph with a text check was dropped
again. The owner is now read by segment, attributes first.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* fix(desktop): draft text checks from text in runs added whole, and name a text element by its own id before its container
Text typed into a run the diff lists whole was never read, so the draft kept
only a presence check on the run, which wrong text also passed. Such text now
marks the nearest element both ends hold, or, when none can take the check,
the added element's own text. The climb also names an element by an
identifier of its own (id, xml:id, xid) before a larger container, so a
headline under a long layer keeps a whole-text check that survives re-split
runs.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* test(desktop): expect the whole-text check where #8988's pairing test meets text checks
On this branch a changed element's text is drafted as `text_equals` on the
element, so the identity pairing test now expects that check, under the
same `[@role=body]` key.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* fix(desktop): keep checks for elements without text beside a text edit, and normalize DOCTYPE system id line endings
A review found two gaps.
- A paragraph's whole-text check covered every element added or removed
inside it, including ones holding no text: an image frame, a line break.
A solver who typed the text but never inserted the image passed. Such an
element now keeps its own presence or absence check.
- The Python reader kept a DOCTYPE system literal's CRLF where TypeScript
normalizes it to LF. Both now report LF.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* test(desktop): import textSyntaxFormat once in the state format tests
main's mixed-case JSON suffix fix and this branch both added the import,
and the merge kept both.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
---------
Co-authored-by: Claude Opus 5.5 <[email protected]>
Fixes#8838.
Admit all symbolic query entry points through the existing resolver budget, preserve cache-hit depth and restore entered frames in finally. Retain independent facts with partial source coverage instead of losing entire scripts to recursive spreads.
Validated 273 focused tests, check:fast/docs:check, real CLI/stdio-MCP projections and complete parent reads, independent public-source oracle, and unchanged full results for the official Electron control and Bitwarden selected preload.
Full technical CI 42308068679 passes, including all eight shards and required aggregates. Greptile reviews this head at 5/5; no completed Codex review of this head is available. The Author readiness-only exception is explicitly authorized; no human-review attestation is added.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Fixes#7540.
Retain every raw dryRun literal at the shared inert token binder before Incur boolean coercion; preserve parser-owned values, original execution, defaults and variadics. Return an actionable typed PARSE_ERROR before setup or environment publication.
Validated 40 focused tests, independent nine-call compiled-CLI before/after configuration controls, check:fast and docs:check.
Full technical CI 42302162266 passes, including all eight shards and required aggregates. Greptile reviews this head at 5/5; no completed Codex review of this head is available. The Author readiness-only exception is explicitly authorized; no human-review attestation is added.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Fixes#7375. Fixes#7608.
Share argument consumption and agent facts between admission and scope checks; reject invalid registrations before settings, backups or runner startup. Preserve missing-value priority, explicit redaction, discovery and valid registration controls.
Validated 33 focused tests (one platform skip), current-main RED and independent compiled-CLI before/after workflows; check:fast passes.
Full technical CI 42297438912 passes, including all eight shards and required aggregates. Greptile reviews this head at 5/5; no completed Codex review is available and the bot reports exhausted quota. The original author readiness declarations remain verbatim; no new human-review attestation is added.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Compare the preceding valid TypeDef coordinate independently for FieldList and MethodList, retaining readable members while leaving affected ownership unknown. Keep the scalar producer and shared owner lookup authoritative across member inspection, metadata inventory and native boundary workflows.
Validated with 577 managed/workflow tests, seven malformed/working ordering vectors, and independent dnfile/pefile raw metadata from a pinned public SystemEvents assembly. Four deliberately changed metadata bytes reproduce both decreases across an invalid row. Six CLI calls and ten stdio MCP tool calls before/after preserve all 458 fields and 195 methods and raw signatures/CIL hashes, keep ordinary results identical across three operations, and retain unresolved filter candidates rather than false absence. Alternative #8868 passes the expanded 206 inspector/ownership controls; converge on this generic implementation.
The original three unchecked human declarations are preserved. The explicitly authorized readiness-only exception applies after technical CI and review; no human-review attestation is added.
Fixes#8861
Full technical CI 42293217445 passes, including all eight shards and required aggregates. Greptile reviews this head at 5/5; Codex cannot complete because the repository review quota is exhausted. The Author readiness-only exception is explicitly authorized; no human-review attestation is added.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Classify a native method-93 decoder window refusal once at the ZIP owner as limit/memory, retaining the fixed 128 MiB cap, native cause and existing corruption/CRC/size/cancellation handling. Derive recovery labels from the cap and share one independent frame fixture.
Validated with 43 reader/translationtests,three actual CLI/stdio MCP cases, a pinned public GLFW payload rewrapped in independently verified ZIP controls, six equivalent CLI calls before/after, static checks and documentation checks. All input bytes remain unchanged, normal extraction matches exactly, and failed outputs are absent. ELF/GPU component refusal-policy gaps remain separate unverified full-format workflows.
The original three unchecked human declarations are preserved. The explicitly authorized readiness-only exception applies after technical CI and review; no human-review attestation is added.
Fixes#8623
Full technical CI 42279406899 passes, including all eight shards and required aggregates. Greptile reviews this head at 5/5; Codex cannot complete because the repository review quota is exhausted. The Author readiness-only exception is explicitly authorized; no human-review attestation is added.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Keep independent Objective-C class, metaclass, category and protocol records when an adopted-protocol list is unreadable. The shared list reader retains readable prefixes through tagged Result failures; consumers preserve partial/unknown semantics with unchanged limits. Advance the source-owned reader commitment to v3 and reject stale v2 snapshot bindings.
Validated with 232 focused tests, current-main and original-proposal RED checks, named provider-result/schema/identity controls, and a verified public GLFW Mach-O paired with independent LLVM/Python byte evidence. Ordinary binary results remain identical; a deliberately mutated count preserves ten rows instead of eight. The final head passes the real macOS CI lane, with twelve CLI tests including both new protocol-count cases over CLI and stdio MCP. Snapshot binding and same-version replay checks pass.
The original unchecked human declarations are preserved. The explicit maintainer readiness exception applies after technical CI and automated review, without adding a human-review attestation.
Fixes#8945
Full technical CI 42263548929 passes, including all eight shards and required aggregates. The Author readiness-only exception is explicitly authorized; no human-review attestation is added.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
* fix(desktop): read the PPA key add-apt-repository writes into Signed-By
On Ubuntu 24.04, add-apt-repository puts the PPA's armored key inside the
deb822 Signed-By field, and REA refused such repositories, so PPAs added the
standard way could not be exported. REA now reads that key by deb822 rules,
verifies the release with it through the same private gpgv keyring as key
files, as the entry's only candidate (as apt does), retains it as armored and
exports it as apt-keys/<name>.asc. A held key that does not verify is refused
with a reason that names no URL.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* test(desktop): check the held-key refusal's facts, not its sentence
The PPA refusal tests now assert that the reason names the sources file
and the release file and no URL, and that export carries that reason,
instead of comparing whole sentences.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
---------
Co-authored-by: Claude Opus 5.5 <[email protected]>
* test(process): align symlink fixture with filesystem checkpoints
* test(frida): assert the CLI spawn Evidence contract
* test(mcp): use an unsupported analysis-view parent
* test(browser): separate source-location line bounds from parity
* fix(dotnet): reject invalid UTF-8 metadata version strings
* test(dotnet): preserve valid multibyte metadata version text
* fix(desktop): recognise JSON scalar files with mixed-case suffixes
* test(setup): use the current registration command in CodeBuddy fixture
* test(cli): split doctor help and timing boundary groups
* style(test): format split MCP doctor boundary groups
Resolve exported SQLite metadata through built-in PRAGMAs on the private connection. Preserve application tables/views, source encoding, nullable rows and the existing resource limits.
Validated with 93 focused tests, current-main RED controls, and actual CLI exports of the public Chinook dataset with native/shipped-reader parity and unchanged source hashes. No live GUI, container or installed agentenv qualification is claimed.
Full technical CI 42247722923 (all eight shards and final aggregate) passed; Greptile review scored 5/5 on the merged head. The original human readiness declarations remain unfilled; the explicitly authorized maintainer readiness exception applies without adding a human-review attestation.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Keep full schemas and canonical analysis unchanged while loading the cross-domain view through existing named discovery. Qualify real recent-tool workflows with token ledgers, latency, CPU, family RSS and retained-follow-up controls.
BREAKING CHANGE: The opt-in hot-cold catalog no longer preloads inspect_analysis_view. Discover it before calling it, batch with its producer when needed, or use the unchanged default full profile.
Preserve per-side capacity diagnostics in shared managed member comparisons and apply repeatable CLI type selection to both source inspections. Keep partial and historical unknown semantics intact, with one typed diagnostic projection and side schema.
Validated with 159 focused tests, final-main CLI/MCP boundary checks, schema validation and official public WPF 8.0.32/8.0.33 assemblies. Full source Evidence remains available through the retained-record MCP workflow. No target execution or budget increase.
Fixes#9021
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
* fix(javascript): apply package type scope to application sources
Application analysis parsed .js sources without their package scope, so an unbound require in a type-module package became a resolved CommonJS dependency. The parent now observes each source's nearest same-container package.json, stopping at node_modules, and the worker parses with that package type as reference imports already do.
Fixes#7323
* fix(javascript): keep package type unknown behind unavailable manifests
The package-type lookup indexed only readable files, so a nearer
package.json whose bytes were unavailable was skipped and an ancestor's
"type" applied to sources it does not own. Index unavailable manifests
by container and directory too, and stop the walk with "unknown" when
one is the nearest owner, as the Next app-context lookup already does.
* fix(javascript): keep Electron preloads outside module package scope
Electron's preload loader ignores a package's "type": "module" and gives
a .js preload `require`; ESM preloads need the .mjs suffix. Applying the
package type to every .js source marked such a preload as an ES module,
so its `require("electron")` edges and IPC pairing disappeared.
A source that requires an Electron module now keeps the context REA
inferred before package scope was applied. Node-loaded sources in a
module scope still drop unbound `require`.
* fix(javascript): admit Electron script loaders only on renderer evidence
The previous commit dropped a module package type whenever a source
required an Electron module. A Node-loaded ES module that requires
Electron, or one whose Electron call goes through a local `require`,
then regained the false require edges the package scope removes.
The parser applies the package type unconditionally again. The worker
re-reads a package-module script that requires an Electron module once
outside the package scope. The source keeps that context only when the
static pass there directly observes a contextBridge exposure or a
renderer IPC operation, which only Electron's renderer-side loaders
provide.
* fix(javascript): check Electron renderer evidence without a second static pass
b54845a6b ran the full static pass on an unscoped parse of every
package-module script that requires Electron, then discarded it when no
renderer operation turned up, so such main-process sources paid for two
full passes.
A candidate must now also name contextBridge or ipcRenderer, which any
renderer evidence needs. The evidence check runs only the static pass's
binding and Electron identity preamble with its exposure and IPC
descriptors. The full static pass runs once, on the chosen parse.
* fix(javascript): confirm Electron renderer evidence with the static pass
The focused evidence check from 4a86bca9b also counted renderer calls in
bundler factory defaults that the bundler convention never runs, which
the static pass skips. Such a default switched a type-module file to
script context, so its unrelated unbound requires became imports again.
The focused check now only screens. It follows the static pass's own
traversal, so a negative result still rules evidence out. A positive
result is confirmed by the static pass's direct findings on the unscoped
parse, and those findings become the file's facts when they hold.
* fix(javascript): confirm renderer evidence before wrapper substitution
f71236400 confirmed a screened source from the static pass's final
contextBridge and IPC findings, without wrapper substitutions. The
wrapper pass replaces a directly observed exposure inside a helper with
a substituted finding, so such a preload lost its imports again.
Before its wrapper pass, the static pass now records whether its source
traversal directly observed a contextBridge exposure or a renderer IPC
operation, and observedElectronRendererCall exposes that. The traversal
still skips bundler factory defaults that never run, and the reported
findings are unchanged.
* fix(javascript): resolve declared preload loaders before source analysis
* fix(javascript): leave preload discovery failures to source analysis
A source that failed during preload discovery threw out of the whole
application analysis, so one oversized source or a per-source timeout in a
module package returned no Evidence. Skip a source that exceeds its worker
limits and end discovery on any other failure except cancellation; source
analysis then meets the same source, records its failure and keeps partial
Evidence as it does without discovery.
* fix(javascript): report preload discovery gaps in Evidence limitations
When preload discovery skipped a source or stopped early, source analysis
could still read that source, so the result looked complete while a .js
preload declared there kept its package scope. Return each discovery gap
with the discovered paths and add it to the analysis limitations, which
both complete and partial Evidence carry.
* feat(process): add a non-executing dry run to capture_process_scenario
Preparation prompts asked an agent to prepare a scenario without running
it but could only suggest the tool that launches the target. Add
dry_run: true so the CLI and MCP resolve the executable, working
directory, observation paths and admission checks, then return planned
Evidence under the rea.process-scenario-plan predicate. Planned Evidence
is not capture Evidence and comparison tools refuse it as a capture.
The preparation prompt now suggests the dry-run call, and the scenario
identity commitments exclude the operation-only field.
Fixes#8398
* fix(process): redact planned sensitive input and share admission
Address review findings on the dry-run contract:
- Redact sensitive input events in planned Evidence and preserve only
their byte length.
- Recheck cancellation after the last awaited admission step.
- Run the sensitive-matcher and filesystem-root admission checks that a
capture performs before launch.
- Assert the prompt advertises the dry_run field without freezing
explanatory wording.
* test(process): hold the final admission step for cancellation
Add a preparation host seam so the regression can cancel while the last
filesystem admission call is pending, then assert the returned
AnalysisCancelledError retains the caller's reason.
* refactor(process): remove the unused plan alias
Keep the plan schema as the single exported contract and satisfy the
unused-export check.
* fix(process): report admission-time aborts as planning cancellation
`prepareProcessScenarioPlan` returned the admission error verbatim, so an
abort that landed while `admitFilesystemObservationRoots` awaited its
`lstat` surfaced as a cancelled `ProcessCaptureError` with
`operation: process_capture` even though no process ran. Convert that
result through the same `cancelled()` projection the neighbouring checks
use, and pin it with a test.
* fix(contracts): keep the process result invariants on the dry-run union
The output schema for `capture_process_scenario` became a union of the
capture and plan shapes, and the union no longer carried the description
that advertised the cross-field invariants to MCP clients — the boundary
test reads `normalized_result.description`. Attach a description to the
union that names both variants and keeps the capture invariants verbatim.
* test(process): exercise the post-admission cancellation check
The previous regression aborted immediately after starting preparation and
released the held admission afterwards. Preparation awaits runtime path
resolution first, so that abort was always observed by the pre-admission
check; deleting the post-admission "if (aborted()) return cancelled();"
still left the test green.
Signal from the admitRoots seam when admission is actually entered, abort
only after that point, then let admission resolve ok. Preparation now has
to reach the check between a successful admission and the ok result, and
the assertion pins both the AnalysisCancelledError tag and the caller's
reason as its cause.
Verified red-green: removing only the post-admission check fails this test
(expected true to be false at prepared.ok) while the other two pass, and
restoring it passes all three.
* test(process): gate the admission seam on acknowledgement
Race the admission acknowledgement against planning's own settlement and a
bounded deadline, and release and join planning in `finally`, so a planning
result that arrives before admission fails the test instead of stalling it
until the suite timeout.
* fix(contracts): hold plans to the scenario refinements and redact inputs
`processScenarioPlanSchema` described the resolved dry-run response but
dropped every cross-field check the input contract applies, so a plan
could advertise out-of-order events, an event past `timeout_ms`, or a
terminal beyond the renderer cell budget while claiming it had been
validated. Extract those checks into `scenarioRefinements` and refine the
plan's scenario with them.
The plan's input event also carried `sensitive` and an optional
`input_bytes`, which accepted a sensitive event that still shipped its
data. Model the public and sensitive forms as separate exact variants:
public input forbids `input_bytes`, sensitive input requires an empty
`data` plus its byte length.
---------
Co-authored-by: 2SinceJuly <[email protected]>
* feat(rizin): select the decoder and map address for one-shot commands
execute_rizin_command and rea execute-rizin-command accept optional
asm_arch, asm_bits, asm_cpu and base_address (CLI --asm-arch,
--asm-bits, --asm-cpu and --base-address). REA passes them to Rizin
before the artifact path as -a, -b, -e asm.cpu= and -m. Rizin 0.9.1
refuses -e asm.arch= and -e asm.bits= at startup, while -a and -b work
in 0.7.4 and 0.9.1. Rizin applies -e settings after -a and -b.
base_address is the address of file offset 0, as in REA's raw-binary
inputs.
REA checks only the spelling, because Rizin owns the arch and CPU
vocabulary and its builds differ. The Homebrew build of Rizin 0.9.1
has no xtensa or riscv plugin; the official static Linux build of
0.7.4 has both. Evidence records each input and the complete argument
list. When any input is set, Evidence also states that REA does not
confirm the decoder or map that Rizin used: Rizin uses another value
for one it cannot apply and exits 0. Rizin 0.9.1 keeps its default and
reports the value on stderr, and Rizin 0.7.4 turned -b 8 for arm into
16 without a message. Calls without these inputs keep their argument
list and Evidence.
The real-provider gate tests/boundary/providers/RizinDecoderInputs.test.ts
checks the effect of each flag through MCP with Rizin's own e asm.*,
omlj and byte reads. It skips when no Rizin is on PATH. It passes with
Rizin 0.9.1 (Homebrew, macOS arm64) and 0.7.4 (official static build,
Linux x86_64).
* docs(rizin): map multi-segment ESP32 images with om
An ESP32-family app or bootloader image holds several segments, each
loaded at its own address, so one base_address cannot map it. The
recipe removes the default map with om-* and adds one om map per
segment, using the segment offsets from the image format or from
esptool image-info. It does not depend on an ESP image reader in REA.
The offsets were checked against esptool 4.12.0 image_info on a
synthetic two-segment ESP32 image. Rizin 0.9.1 and 0.7.4 mapped its
segments at their load addresses. Rizin 0.7.4, whose static Linux
build has the xtensa plugin, decoded the IRAM segment as nop.n, nop.n,
ret.n and nop.n.
* fix(rizin): read the file at base_address when a loader would claim it
On Rizin 0.9.1, base_address could read 0xff at the mapped address with
exit 0. -m maps the file, but Rizin still runs its format loaders, and
a loose one can claim a raw image. Rizin 0.9.1's AVR loader matches
when bytes 1 and 3 look like rjmp and turns io.va off, so reads use
file offsets. The BIOS loader, in 0.7.4 and 0.9.1, matches a 64 KiB
file with a jmp 16 bytes from its end and maps it at another address.
An acceptance run on a raw ESP32-S3 segment hit the first case.
With base_address, REA now passes -n together with -m. Rizin then skips
the loaders and maps the whole file at the address: rizin.c calls
rz_io_map_new at the -m address and leaves io.va at its default. The
input therefore means a raw image, and Rizin loads no sections, symbols
or entry point, even from an ELF file. Evidence states this in its own
limitation, and the decoder limitation no longer covers the map.
The real-provider gate adds an AVR-like and a BIOS-like image and reads
below the base, at it, at the end of the image and past it. Without -n
it fails on Rizin 0.9.1. It passes with Rizin 0.9.1 (Homebrew, macOS
arm64) and 0.7.4 (official static build, Linux x86_64). The ESP om
recipe now starts with e io.va=true, because reads use the maps only
while io.va is on.
* fix(rizin): advertise name patterns that compile in v mode
The asm_arch and asm_cpu pattern ended its character class with an
unescaped '-', which JavaScript's v (unicodeSets) mode rejects, so
clients compiling advertised patterns in that mode could not use
execute_rizin_command's schema. Escaping it keeps the same names valid
in every mode, as main's own patterns write it.
* fix(rizin): state what the decoder and map inputs leave unconfirmed
The base_address limitation reported a map REA never observes and left
out that -n keeps Rizin's default arch, bits and byte order. The asm_*
limitation implied Rizin replaces a bad CPU, but e asm.cpu echoes any
value. Compact schemas drop property descriptions, so the tool
description now carries the raw-loading warning. The real-provider
gate checks that asm_cpu changes a decode, and a base_address-only call.
* docs(rizin): record the gate's full run on Rizin 0.7.4
* test(rizin): follow the test guide in the decoder input tests
Focused tests resolve execute_rizin_command from its contract family,
the harness owns the temporary directory before setup can fail, and
guidance is compared with its owner, not with sentences from it.
* feat(domain): guide agents from neighboring-tool argument mistakes (#2146)
* feat(mcp): guide agents from argument-name mistakes in MCP tools (#2146)
* fix(domain): suppress argument replacement when expected field is present (#2146)
Suppress argument replacement guidance when the expected field is already present in the call input, preventing invalid suggestions such as advising procedure when procedure was already provided. Defrost explanatory prose assertions in tests and bind test cases to production toolContract schemas.
* test(domain): use named contract families and decouple guidance prose in tests (#2146)
Import named contract families (OFFICIAL_TOOL_CONTRACTS, APPLICATION_TOOL_CONTRACTS, ENHANCED_TOOL_CONTRACTS) instead of the full tool catalog per docs/testing.md. Replace hardcoded explanatory message and remediation strings with dynamic references to the owning guidance constants.
* test(domain): restore independent replacement field assertions (#2146)
Keep exact checks for expected replacement field names ('procedure', 'pattern', 'address', 'query') independent of the owning guidance helper per docs/testing.md, while retaining dynamic validation for message and remediation prose.
morluto