Trim eager read tool schemas

* docs(agents): codify recurring review blockers in AGENTS.md

Capture the high-frequency classes that drive multi-round maintainer reviews
(stale bases, path canonicalization, secret handling, partial setup, missing
regression tests, scope honesty) so coding agents fix them before first review.

Refs #856

* docs(agents): tighten security and atomic-write review guidance

Require open-time containment binding instead of pre-open EvalSymlinks, and
require temp-file then atomic replace rather than exclusive create alone.

Refs #856

* docs(agents): require traversal-wide containment for no-follow APIs

Final-component-only O_NOFOLLOW is insufficient on Linux and reparse
points need platform protections. Align the Security edges guidance
with the full-path containment requirement from review.

AGENTS.md

@@ -67,3 +67,39 @@ requirements. If a related check fails, fix it. If a required check cannot run
 because of the environment or fails for an unrelated external reason, report
 the exact failure and obtain maintainer direction rather than silently ignoring
 it.
+
+## 4. Common Review Blockers
+
+These classes drive multi-round reviews. Fix them before requesting review:
+
+- **Fresh base:** Rebase onto the current PR base (`main` or stacked target)
+  before review. A stale head that rolls back mainline commits is a hard
+  blocker, not a merge-time detail. Resolve conflicts by keeping upstream
+  behavior unless the PR intentionally changes it.
+- **Platform truth:** Code and tests must pass on Linux, macOS, and Windows.
+  Compare paths after canonicalization (physical/`EvalSymlinks` form); never
+  assert raw `t.TempDir()` spellings (`/var` vs `/private/var` on macOS, short
+  paths on Windows). Prefer one cross-platform function with small `GOOS`
+  checks over duplicated helpers that drift.
+- **Security edges:** Keep secrets out of argv, env dumps, and logs. Redact
+  success and error paths (including stderr). Fail closed on ownership, lease,
+  and permission checks. Do not rely on pre-open path resolution
+  (`EvalSymlinks` then open) for containment: that is a check-to-use race.
+  Bind containment at open/use time with rooted or handle-relative,
+  traversal-resistant APIs. If a no-follow API is used, apply it to every
+  traversed component and enforce the platform's reparse-point protections;
+  final-component-only no-follow is insufficient. On multi-step setup, roll back
+  only what this run created; never destroy pre-existing resources you did
+  not create; never report success when cleanup or unlock failed.
+- **Atomic shared state:** Write a complete temporary file, then atomically
+  replace the destination so concurrent readers never see a partial write.
+  Exclusive create or a write lock alone is not enough for readers. Serialize
+  the full read-modify-write sequence for lockfiles and shared stores.
+- **Tests match the claim:** Every behavior or security-boundary change needs a
+  regression test, including the failure path. Path-sensitive logic needs at
+  least one non-Linux path case (or a hermetic fake that exercises the same
+  normalization).
+- **Honest scope:** PR description, help text, and comments must match what
+  shipped. Wire advertised entry points or shrink the claim. Do not bundle
+  unrelated fixes. Do not widen security allowlists by name alone; check
+  classification and side effects too.

internal/agent/additional_permissions_validation_test.go

@@ -127,9 +127,7 @@ func TestExecuteToolCallMissingAdditionalPermissionsErrorIsActionable(t *testing
 	if !strings.Contains(result.Output, `{"network": {"enabled": true}}`) {
 		t.Fatalf("output = %q, want a concrete additional_permissions example", result.Output)
 	}
-	if !strings.Contains(result.Output, "omit sandbox_permissions") {
-		t.Fatalf("output = %q, want guidance that the flag can simply be omitted", result.Output)
-	}
+	assertNamesOmitRemedy(t, "missing additional_permissions", result.Output)
 }
 
 // A well-formed additional_permissions payload must still reach the normal
@@ -165,6 +163,123 @@ func TestExecuteToolCallStillPromptsForValidAdditionalPermissions(t *testing.T)
 	}
 }
 
+// runShellCall drives one shell tool call through the real registry (and so
+// through Registry.RunWithOptions) exactly as the agent loop would.
+func runShellCall(t *testing.T, root string, arguments string) ToolResult {
+	t.Helper()
+	registry := tools.NewRegistry()
+	registry.Register(tools.NewScopedExecCommandTool(root, nil, nil))
+	result, err := executeToolCall(
+		context.Background(),
+		registry,
+		ToolCall{ID: "c1", Name: "exec_command", Arguments: arguments},
+		PermissionModeUnsafe,
+		Options{Cwd: root},
+	)
+	if err != nil {
+		t.Fatalf("executeToolCall: %v", err)
+	}
+	return result
+}
+
+// TestAdditionalPermissionsRejectionsConverge is the regression guard for
+// #845/#847. A model that attaches a permission object to a command needing
+// none used to be told only "set sandbox_permissions", which pushed it into the
+// OTHER invalid shape (flag set, payload empty) and its "include a permission"
+// error pushed it straight back — the reported loop that burned turns on
+// `git branch --show-current`.
+//
+// The contract this pins is convergence, not tolerance: both shapes are still
+// REJECTED (nothing here grants access), but each rejection must name the one
+// follow-up call that works, and that follow-up must actually succeed.
+func TestAdditionalPermissionsRejectionsConverge(t *testing.T) {
+	root := t.TempDir()
+	const readOnly = `"cmd":"git --version"`
+
+	// Step 1 of the reported loop: a permission object with no flag.
+	first := runShellCall(t, root, `{`+readOnly+`,"additional_permissions":{"network":{"enabled":true}}}`)
+	if first.Status != tools.StatusError {
+		t.Fatalf("unflagged additional_permissions status = %q, want error (rejecting is correct)", first.Status)
+	}
+	assertNamesOmitRemedy(t, "unflagged additional_permissions", first.Output)
+
+	// Step 2 of the reported loop: the model keeps the flag but empties the
+	// payload. This must not be the dead end it was.
+	second := runShellCall(t, root, `{`+readOnly+`,"sandbox_permissions":"with_additional_permissions"}`)
+	if second.Status != tools.StatusError {
+		t.Fatalf("flag without payload status = %q, want error", second.Status)
+	}
+	assertNamesOmitRemedy(t, "flag without payload", second.Output)
+
+	// The call both errors point at must actually work.
+	converged := runShellCall(t, root, `{`+readOnly+`}`)
+	if converged.Status != tools.StatusOK {
+		t.Fatalf("the call both errors instruct the model to make failed: status=%q output=%q", converged.Status, converged.Output)
+	}
+}
+
+// assertNamesOmitRemedy pins the actionable half of the error text: the message
+// must spell out that BOTH fields can simply be dropped. "Set the flag" or
+// "include a permission" alone is what made the two errors chase each other.
+func assertNamesOmitRemedy(t *testing.T, label string, output string) {
+	t.Helper()
+	if !strings.Contains(output, "omit both sandbox_permissions and additional_permissions") {
+		t.Fatalf("%s error = %q, want it to say to omit both fields entirely", label, output)
+	}
+	if !strings.Contains(output, "retry") {
+		t.Fatalf("%s error = %q, want it to tell the model to retry", label, output)
+	}
+}
+
+// The empty-payload rejection carries the same remedy. It is pinned on the
+// validator directly because normalizeNoopAdditionalPermissions absorbs the
+// empty payload before executeToolCall reaches this branch;
+// inlineAdditionalPermissionsProfile is still the shared source of truth that
+// buildPermissionEvent and the grant path re-run, so its text must converge too.
+func TestEmptyAdditionalPermissionsErrorNamesOmitRemedy(t *testing.T) {
+	args := map[string]any{
+		"cmd":                 "git --version",
+		"sandbox_permissions": string(tools.SandboxPermissionsWithAdditionalPermissions),
+		"additional_permissions": map[string]any{
+			"network":     map[string]any{"enabled": false},
+			"file_system": map[string]any{"read": []any{}, "write": []any{}},
+		},
+	}
+	_, requested, err := inlineAdditionalPermissionsProfile(args, t.TempDir())
+	if !requested {
+		t.Fatal("requested = false, want true for an explicit with_additional_permissions call")
+	}
+	if err == nil {
+		t.Fatal("an empty additional_permissions payload was accepted; it must still be rejected")
+	}
+	assertNamesOmitRemedy(t, "empty payload", err.Error())
+}
+
+// A provider that materializes every optional property may send
+// "additional_permissions": null. That grants nothing and carries no flag, so it
+// is the field omitted — rejecting it stalled an ordinary command behind an
+// error the model could not act on.
+func TestExecuteToolCallTreatsNullAdditionalPermissionsAsOmitted(t *testing.T) {
+	result := runShellCall(t, t.TempDir(), `{"cmd":"git --version","additional_permissions":null}`)
+	if result.Status != tools.StatusOK {
+		t.Fatalf("null additional_permissions was rejected: status=%q output=%q", result.Status, result.Output)
+	}
+}
+
+// A null payload alongside the flag must still be rejected, and must land on
+// the same error as an absent payload rather than being quietly downgraded.
+func TestExecuteToolCallRejectsFlaggedNullAdditionalPermissions(t *testing.T) {
+	result := runShellCall(t, t.TempDir(),
+		`{"cmd":"git --version","sandbox_permissions":"with_additional_permissions","additional_permissions":null}`)
+	if result.Status != tools.StatusError {
+		t.Fatalf("flagged null status = %q, want error", result.Status)
+	}
+	if !strings.Contains(result.Output, "no additional_permissions object was provided") {
+		t.Fatalf("output = %q, want the missing-payload error", result.Output)
+	}
+	assertNamesOmitRemedy(t, "flagged null", result.Output)
+}
+
 func TestNormalizeProviderExpandedEmptyAdditionalPermissions(t *testing.T) {
 	for _, mode := range []string{
 		string(tools.SandboxPermissionsUseDefault),

internal/agent/loop.go

@@ -656,6 +656,13 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 		// between tool_results breaks strict provider replay) — same after-batch
 		// rationale as turnRequestedModel above.
 		var changedFilesThisBatch []string
+		// Images produced by tools this turn, held until every tool_result is
+		// recorded. They travel as user messages, and a user message between two
+		// tool_results breaks strict provider replay — Anthropic coalesces them
+		// into one user block list and requires the tool_result blocks first, so
+		// interleaving yields [tool_result, text, image, tool_result] and a 400.
+		// Same reason the self-correction feedback below is deferred.
+		var toolImageMessages []zeroruntime.Message
 		// Parallel read-ahead state: results for calls[precomputedStart:precomputedEnd]
 		// executed concurrently, consumed strictly in order below.
 		var precomputed []precomputedToolResult
@@ -709,6 +716,16 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 				Content:    toolResult.Output,
 				ToolCallID: toolResult.ToolCallID,
 			})
+			// Images ride a following USER message rather than the tool result
+			// above. Every provider drops images on a tool-role message —
+			// Anthropic's tool_result content is a string, Gemini's is a
+			// functionResponse, and OpenAI guards its image parts to the user role
+			// — so attaching them there would silently deliver nothing. A separate
+			// message also keeps the one-tool-result-per-tool-call pairing intact,
+			// which the providers validate.
+			if imageMessage, ok := toolResultImageMessage(toolResult); ok {
+				toolImageMessages = append(toolImageMessages, imageMessage)
+			}
 
 			// A tool may demand the run ABORT — a canceled/timed-out ask_user prompt
 			// returns context.Canceled rather than fabricating a headless answer. Stop
@@ -719,11 +736,13 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 			}
 			if abortErr != nil {
 				messages = appendAbortedToolResults(messages, collected.ToolCalls[index+1:])
+				messages = append(messages, toolImageMessages...)
 				result.Messages = copyMessages(messages)
 				return result, abortErr
 			}
 			if stopReason := stopReasonFromToolResult(toolResult); stopReason != "" {
 				messages = appendAbortedToolResults(messages, collected.ToolCalls[index+1:])
+				messages = append(messages, toolImageMessages...)
 				result.FinalAnswer = toolResult.Output
 				result.StopReason = stopReason
 				result.Messages = copyMessages(messages)
@@ -747,6 +766,7 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 				// messages stay valid for a strict provider replay (Anthropic
 				// rejects a tool_use with no answering tool_result).
 				messages = appendAbortedToolResults(messages, collected.ToolCalls[index+1:])
+				messages = append(messages, toolImageMessages...)
 				result.FinalAnswer = toolFailureStopAnswer(call.Name, outcome.Count)
 				result.Messages = copyMessages(messages)
 				return result, nil
@@ -765,6 +785,10 @@ func Run(ctx context.Context, prompt string, provider Provider, options Options)
 				postEditDiagnostics.enqueue(ctx, toolResult.ChangedFiles)
 			}
 		}
+		// Every tool_result for this turn is now recorded, including aborted
+		// placeholders, so the images can follow without splitting them.
+		messages = append(messages, toolImageMessages...)
+		toolImageMessages = nil
 
 		// Run post-edit self-correction once over the union of files this turn
 		// changed, then append any feedback after every tool_result is recorded so
@@ -1436,6 +1460,7 @@ func executeToolCall(ctx context.Context, registry *tools.Registry, call ToolCal
 		Output:          result.Output,
 		Truncated:       result.Truncated,
 		Meta:            result.Meta,
+		Images:          result.Images,
 		Redacted:        result.Redacted,
 		ChangedFiles:    result.ChangedFiles,
 		ChangeSummaries: result.ChangeSummaries,
@@ -2763,7 +2788,16 @@ func shellCommandAdditionalPermissionsRequested(args map[string]any) bool {
 // access and malformed payloads are left untouched so validation stays closed.
 func normalizeNoopAdditionalPermissions(args map[string]any, basePath string, engine *sandbox.Engine) {
 	raw, exists := args["additional_permissions"]
-	if !exists || raw == nil {
+	if !exists {
+		return
+	}
+	// An explicit null is the field omitted, not a request: drop the key so the
+	// rest of the pipeline sees the same args a model that omitted it would send.
+	// sandbox_permissions is deliberately left alone — a flagged null must reach
+	// the same "no additional_permissions object was provided" error as a flagged
+	// absent field rather than being silently downgraded to a different outcome.
+	if raw == nil {
+		delete(args, "additional_permissions")
 		return
 	}
 	data, err := json.Marshal(raw)
@@ -2794,21 +2828,41 @@ func normalizeNoopAdditionalPermissions(args map[string]any, basePath string, en
 	}
 }
 
+// omitAdditionalPermissionsHint is the one instruction that ends the retry loop
+// in #845/#847. Every additional_permissions rejection must carry it: the model
+// reaches these errors by attaching a permission object to a command that needs
+// none, so "omit it and retry" is the answer in practice, and an error that
+// only explains how to build a VALID permission object sends the model round
+// the loop again with a different invalid shape.
+const omitAdditionalPermissionsHint = "If this command does not need extra sandbox access — ordinary " +
+	"read-only commands such as `git branch --show-current` do not — omit both sandbox_permissions and " +
+	"additional_permissions entirely and retry; that is the fix in almost every case."
+
 func inlineAdditionalPermissionsProfile(args map[string]any, basePath string) (sandbox.RequestPermissionProfile, bool, error) {
 	if !shellCommandAdditionalPermissionsRequested(args) {
-		if _, exists := args["additional_permissions"]; exists {
-			return sandbox.RequestPermissionProfile{}, false, fmt.Errorf("additional_permissions requires sandbox_permissions set to %q", tools.SandboxPermissionsWithAdditionalPermissions)
+		// An explicit null grants nothing and carries no flag, so it is
+		// indistinguishable from an omitted field; rejecting it would be a pure
+		// false positive against providers that materialize every optional
+		// property.
+		if raw, exists := args["additional_permissions"]; exists && raw != nil {
+			return sandbox.RequestPermissionProfile{}, false, fmt.Errorf(
+				"additional_permissions was provided without sandbox_permissions set to %q, so it cannot be applied. %s "+
+					"Only if the command genuinely needs file or network access the sandbox denies, resend with "+
+					`sandbox_permissions: %q AND a non-empty additional_permissions, for example {"network": {"enabled": true}}`,
+				tools.SandboxPermissionsWithAdditionalPermissions,
+				omitAdditionalPermissionsHint,
+				tools.SandboxPermissionsWithAdditionalPermissions)
 		}
 		return sandbox.RequestPermissionProfile{}, false, nil
 	}
 	raw, ok := args["additional_permissions"]
 	if !ok || raw == nil {
 		return sandbox.RequestPermissionProfile{}, true, fmt.Errorf(
-			"sandbox_permissions was set to %q but no additional_permissions object was provided. "+
-				`Include one, for example additional_permissions: {"network": {"enabled": true}} or `+
-				`{"file_system": {"write": ["/path"]}}. If this command does not need elevated permissions, `+
-				"omit sandbox_permissions entirely and retry",
-			tools.SandboxPermissionsWithAdditionalPermissions)
+			"sandbox_permissions was set to %q but no additional_permissions object was provided. %s "+
+				`Otherwise include one, for example additional_permissions: {"network": {"enabled": true}} or `+
+				`{"file_system": {"write": ["/path"]}}`,
+			tools.SandboxPermissionsWithAdditionalPermissions,
+			omitAdditionalPermissionsHint)
 	}
 	data, err := json.Marshal(raw)
 	if err != nil {
@@ -2824,8 +2878,10 @@ func inlineAdditionalPermissionsProfile(args map[string]any, basePath string) (s
 	}
 	if normalized.Empty() {
 		return sandbox.RequestPermissionProfile{}, true, fmt.Errorf(
-			"additional_permissions must include at least one of network or file_system, for example " +
-				`{"network": {"enabled": true}} or {"file_system": {"write": ["/path"]}}`)
+			"additional_permissions granted nothing, so it cannot be applied. %s "+
+				`Otherwise include a real permission, for example {"network": {"enabled": true}} or `+
+				`{"file_system": {"write": ["/path"]}}`,
+			omitAdditionalPermissionsHint)
 	}
 	grantProfile, err := sandbox.RequestPermissionGrantProfile(normalized)
 	if err != nil {
@@ -3258,3 +3314,37 @@ func copyMessages(messages []Message) []Message {
 	}
 	return copied
 }
+
+// toolResultImageMessage builds the user message that carries a tool's images
+// to the model, or reports false when the tool produced none.
+//
+// The text names the tool so the model can tell which call an image came from
+// when several ran in one turn — the images arrive detached from their tool
+// result, so nothing else associates them.
+func toolResultImageMessage(result ToolResult) (zeroruntime.Message, bool) {
+	images := make([]zeroruntime.ImageBlock, 0, len(result.Images))
+	for _, image := range result.Images {
+		if len(image.Data) == 0 {
+			continue
+		}
+		mediaType := zeroruntime.NormalizeImageMediaType(image.MediaType)
+		if mediaType == "" {
+			// Outside the provider allow-list. Dropping it beats sending bytes a
+			// provider will reject and failing the whole turn.
+			continue
+		}
+		images = append(images, zeroruntime.ImageBlock{MediaType: mediaType, Data: image.Data})
+	}
+	if len(images) == 0 {
+		return zeroruntime.Message{}, false
+	}
+	label := result.Name
+	if label == "" {
+		label = "tool"
+	}
+	return zeroruntime.Message{
+		Role:    zeroruntime.MessageRoleUser,
+		Content: "Image output from " + label + ":",
+		Images:  images,
+	}, true
+}

internal/agent/system_prompt.md

@@ -30,11 +30,13 @@ work.
    read-before-edit discipline: inspect the target file and nearby callers,
    tests, or config before you modify behavior. Never edit a file you have not
    read.
-2. **Plan.** For multi-step work, call update_plan with an ordered checklist and
-   keep it current at meaningful milestones. Mark completed work and the next
-   in-progress step together when practical; do not spend separate model turns
-   updating the plan after every file or command. Keep at most one item
-   in_progress. Skip the plan for trivial one-step tasks.
+2. **Plan.** Use update_plan for implementation or investigation that has at
+   least three meaningful dependent steps, then keep it current at milestones.
+   Mark completed work and the next in-progress step together when practical;
+   do not spend separate turns updating it after every file or command. Keep at
+   most one item in_progress. Skip plans for simple lookups, explanations, code
+   navigation, and other short tasks. Never create a plan after the work is
+   already complete merely to describe what you did.
 3. **Implement.** Make focused changes that match the surrounding code's style,
    naming, and conventions. Prefer the smallest change that fully solves the
    problem. Avoid broad refactors, unrelated rewrites, dependency churn, and
@@ -49,9 +51,13 @@ work.
 ## Editing discipline
 
 - Choose the narrowest tool that safely accomplishes the step. Prefer native
-  file tools - read_file, list_directory, glob, grep, write_file, edit_file,
-  apply_patch - over shelling out to cat/sed/awk/python for file operations.
+  file tools - read_file, read_minified_file, list_directory, glob, grep,
+  write_file, edit_file, apply_patch - over shelling out to
+  cat/sed/awk/python for file operations.
   They are safer, reviewable, and produce clean diffs.
+- Prefer read_minified_file when initially exploring source code; it preserves
+  code while removing comments and redundant whitespace. Use read_file when
+  exact text, comments, or line numbers are needed.
 - Keep edits focused and reviewable. A single patch may update several related
   files when they form one coherent change; do not hide unrelated edits in a
   bulk shell or script rewrite.

internal/agent/system_prompt_models.go

@@ -71,5 +71,6 @@ const geminiPromptAddendum = `<model_guidance>
   equivalent shell commands; they are safer and produce cleaner diffs.
 - Be concise and concrete. When you run a shell command with side effects, state
   in one short clause why it is needed.
-- Use update_plan for any multi-step task and keep it current.
+- Use update_plan only for work with at least three meaningful dependent steps;
+  skip it for simple lookups and never create one after the work is complete.
 </model_guidance>`

internal/agent/tool_result_images_test.go

@@ -0,0 +1,207 @@
+package agent
+
+import (
+	"context"
+	"fmt"
+	"strings"
+	"testing"
+
+	"github.com/Gitlawb/zero/internal/tools"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+// imageTool returns a tool result carrying an image, the way a screenshot tool
+// should be able to.
+type imageTool struct {
+	media string
+	data  []byte
+}
+
+func (imageTool) Name() string        { return "capture" }
+func (imageTool) Description() string { return "Captures an image." }
+func (imageTool) Parameters() tools.Schema {
+	return tools.Schema{Type: "object", Properties: map[string]tools.PropertySchema{}}
+}
+func (imageTool) Safety() tools.Safety {
+	return tools.Safety{Permission: tools.PermissionAllow}
+}
+func (t imageTool) Run(context.Context, map[string]any) tools.Result {
+	return tools.Result{
+		Status: tools.StatusOK,
+		Output: "Captured a screenshot.",
+		Images: []zeroruntime.ImageBlock{{MediaType: t.media, Data: t.data}},
+	}
+}
+
+// A tool that produces an image must get that image in front of the model.
+//
+// It cannot ride the tool-result message: every provider drops images there.
+// Anthropic maps a tool result to a tool_result block with string content,
+// Gemini to a functionResponse, and OpenAI guards its image parts to the user
+// role explicitly. So the image has to arrive as a following user message,
+// which is also the only shape that keeps one tool result per tool call.
+func TestRunDeliversToolResultImagesToTheModel(t *testing.T) {
+	registry := tools.NewRegistry()
+	registry.Register(imageTool{media: "image/png", data: []byte("\x89PNG\r\n\x1a\nfake")})
+
+	provider := &mockProvider{
+		turns: [][]zeroruntime.StreamEvent{
+			{
+				{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "call_1", ToolName: "capture"},
+				{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "call_1", ArgumentsFragment: `{}`},
+				{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "call_1"},
+				{Type: zeroruntime.StreamEventDone},
+			},
+			{
+				{Type: zeroruntime.StreamEventText, Content: "I can see it."},
+				{Type: zeroruntime.StreamEventDone},
+			},
+		},
+	}
+
+	result, err := Run(context.Background(), "screenshot please", provider, Options{
+		Registry: registry,
+		MaxTurns: 2,
+	})
+	if err != nil {
+		t.Fatalf("Run: %v", err)
+	}
+
+	var carrier *zeroruntime.Message
+	for index := range result.Messages {
+		if len(result.Messages[index].Images) > 0 {
+			carrier = &result.Messages[index]
+			break
+		}
+	}
+	if carrier == nil {
+		t.Fatal("no message carried the tool's image; the model never saw it")
+	}
+	if carrier.Role != zeroruntime.MessageRoleUser {
+		t.Errorf("image rode a %q message; every provider only serializes images on the user role", carrier.Role)
+	}
+	if got := carrier.Images[0].MediaType; got != "image/png" {
+		t.Errorf("media type = %q, want image/png", got)
+	}
+
+	// The tool-result pairing must be untouched: one tool result per tool call.
+	toolResults := 0
+	for _, message := range result.Messages {
+		if message.Role == zeroruntime.MessageRoleTool {
+			toolResults++
+		}
+	}
+	if toolResults != 1 {
+		t.Errorf("got %d tool-result messages for 1 tool call", toolResults)
+	}
+}
+
+// textTool is a second, image-free tool so a turn can carry two calls.
+type textTool struct{}
+
+func (textTool) Name() string        { return "note" }
+func (textTool) Description() string { return "Notes something." }
+func (textTool) Parameters() tools.Schema {
+	return tools.Schema{Type: "object", Properties: map[string]tools.PropertySchema{}}
+}
+func (textTool) Safety() tools.Safety { return tools.Safety{Permission: tools.PermissionAllow} }
+func (textTool) Run(context.Context, map[string]any) tools.Result {
+	return tools.Result{Status: tools.StatusOK, Output: "noted"}
+}
+
+// Tool results for one assistant turn must stay contiguous. The image messages
+// are user-role, and a user message between two tool_results breaks strict
+// provider replay: Anthropic coalesces consecutive user content into one block
+// list and requires tool_result blocks first, so interleaving produces
+// [tool_result, text, image, tool_result] and the request is rejected.
+//
+// The first version of this feature appended each image immediately after its
+// own tool result, which is exactly that shape. A single-tool-call test passed
+// and said nothing about it.
+func TestRunKeepsToolResultsContiguousWhenAToolReturnsAnImage(t *testing.T) {
+	registry := tools.NewRegistry()
+	registry.Register(imageTool{media: "image/png", data: []byte("\x89PNG\r\n\x1a\nfake")})
+	registry.Register(textTool{})
+
+	provider := &mockProvider{turns: [][]zeroruntime.StreamEvent{
+		{
+			{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "c1", ToolName: "capture"},
+			{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "c1", ArgumentsFragment: `{}`},
+			{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "c1"},
+			{Type: zeroruntime.StreamEventToolCallStart, ToolCallID: "c2", ToolName: "note"},
+			{Type: zeroruntime.StreamEventToolCallDelta, ToolCallID: "c2", ArgumentsFragment: `{}`},
+			{Type: zeroruntime.StreamEventToolCallEnd, ToolCallID: "c2"},
+			{Type: zeroruntime.StreamEventDone},
+		},
+		{{Type: zeroruntime.StreamEventText, Content: "done"}, {Type: zeroruntime.StreamEventDone}},
+	}}
+
+	result, err := Run(context.Background(), "two calls", provider, Options{Registry: registry, MaxTurns: 2})
+	if err != nil {
+		t.Fatalf("Run: %v", err)
+	}
+
+	// Two separate ifs rather than a switch: a tool-role message that also
+	// carried images would match only the first arm of a switch, so a regression
+	// that put the image back on the tool result would read as "no image at all"
+	// instead of naming what actually happened.
+	type toolResult struct {
+		index int
+		id    string
+	}
+	var toolResults []toolResult
+	imageIndex := -1
+	for index, message := range result.Messages {
+		if message.Role == zeroruntime.MessageRoleTool {
+			toolResults = append(toolResults, toolResult{index: index, id: message.ToolCallID})
+		}
+		if len(message.Images) > 0 {
+			if imageIndex >= 0 {
+				t.Fatalf("image delivered twice, at %d and %d; recorded %s", imageIndex, index, messageShape(result.Messages))
+			}
+			imageIndex = index
+		}
+	}
+	if imageIndex < 0 {
+		t.Fatalf("the image never reached the model; recorded %s", messageShape(result.Messages))
+	}
+	if len(toolResults) != 2 {
+		t.Fatalf("got %d tool-result messages for 2 tool calls; recorded %s", len(toolResults), messageShape(result.Messages))
+	}
+	if toolResults[0].id != "c1" || toolResults[1].id != "c2" {
+		t.Errorf("tool results answered %q then %q, want c1 then c2", toolResults[0].id, toolResults[1].id)
+	}
+	// The contract Anthropic enforces: every tool_result block has to come before
+	// any other block in the user message it lands in. Anything wedged between
+	// the two results breaks that, which is the 400 this PR exists to avoid.
+	if toolResults[1].index != toolResults[0].index+1 {
+		t.Errorf("tool results at %d and %d are not adjacent; %q message at %d splits them; recorded %s",
+			toolResults[0].index, toolResults[1].index, result.Messages[toolResults[0].index+1].Role,
+			toolResults[0].index+1, messageShape(result.Messages))
+	}
+	// Follows the last tool result, deliberately NOT pinned to exactly last+1.
+	// appendUserBlocks coalesces consecutive user-role content, so
+	// [tool_result, tool_result, text, image] still maps to tool_result blocks
+	// first and replays fine; pinning would redden that valid shape for nothing.
+	if imageIndex <= toolResults[1].index {
+		t.Errorf("image message at %d does not follow the last tool result at %d; recorded %s",
+			imageIndex, toolResults[1].index, messageShape(result.Messages))
+	}
+}
+
+// messageShape renders the conversation as a compact role list so a failure says
+// what the ordering actually was instead of dumping whole messages.
+func messageShape(messages []zeroruntime.Message) string {
+	parts := make([]string, 0, len(messages))
+	for index, message := range messages {
+		part := fmt.Sprintf("%d:%s", index, message.Role)
+		if message.ToolCallID != "" {
+			part += "(" + message.ToolCallID + ")"
+		}
+		if len(message.Images) > 0 {
+			part += "+image"
+		}
+		parts = append(parts, part)
+	}
+	return "[" + strings.Join(parts, " ") + "]"
+}

internal/agent/types.go

@@ -68,8 +68,11 @@ type ToolResult struct {
 	Output     string
 	// Truncated reports that the tool's model-visible output omitted content.
 	// The full result may be recoverable through Meta["spill_path"].
-	Truncated    bool
-	Meta         map[string]string
+	Truncated bool
+	Meta      map[string]string
+	// Images the tool produced, delivered to the model as a following user
+	// message rather than on this result. See tools.Result.Images.
+	Images       []zeroruntime.ImageBlock
 	Redacted     bool
 	ChangedFiles []string
 	// ChangeSummaries are non-selectable generated-tree summaries emitted by

internal/cli/deferred_wiring_test.go

@@ -79,8 +79,8 @@ func TestRegisterToolSearchIfEligibleSkipsWhenThresholdZero(t *testing.T) {
 
 func TestDeferredEligibleCountIncludesOptionalBuiltins(t *testing.T) {
 	registry := newCoreRegistry(t.TempDir())
-	if got := deferredEligibleCount(registry, agent.PermissionModeAuto, nil, nil); got < 3 {
-		t.Fatalf("deferredEligibleCount(core) = %d, want at least 3 optional built-ins", got)
+	if got := deferredEligibleCount(registry, agent.PermissionModeAuto, nil, nil); got < 2 {
+		t.Fatalf("deferredEligibleCount(core) = %d, want at least 2 optional built-ins", got)
 	}
 }
 

internal/cli/mcp_config.go

@@ -308,6 +308,34 @@ func runMCPCheck(ctx context.Context, args []string, stdout io.Writer, stderr io
 	}
 	defer closeMCPRuntime(stderr, mcpRuntime)
 
+	// A connect failure does not come back as err. RegisterTools is deliberately
+	// best-effort so one unreachable server cannot stop Zero launching: it records
+	// the failure on the runtime and returns nil. That is right for startup and
+	// wrong to ignore here, because this is the command a user runs precisely when
+	// a server is misbehaving, and reporting "is reachable" for a server that never
+	// started sends them looking somewhere else.
+	for _, skipped := range mcpRuntime.Skipped() {
+		if skipped.Name != serverName {
+			continue
+		}
+		message := fmt.Sprintf("MCP server %s is not reachable", serverName)
+		if skipped.Err != nil {
+			message = fmt.Sprintf("%s: %s", message, redaction.ErrorMessage(skipped.Err, redaction.Options{}))
+		}
+		if options.json {
+			payload := struct {
+				ServerName string `json:"serverName"`
+				Status     string `json:"status"`
+				Error      string `json:"error,omitempty"`
+			}{ServerName: serverName, Status: "unreachable", Error: message}
+			if err := writePrettyJSON(stdout, redaction.RedactValue(payload, redaction.Options{})); err != nil {
+				return exitCrash
+			}
+			return exitCrash
+		}
+		return writeAppError(stderr, message, exitCrash)
+	}
+
 	items := mcpToolList(registry)
 	if options.json {
 		payload := struct {

internal/cli/mcp_trust_edge_test.go

@@ -6,6 +6,7 @@ package cli
 
 import (
 	"bytes"
+	"encoding/json"
 	"errors"
 	"os"
 	"path/filepath"
@@ -219,3 +220,81 @@ func TestRunMCPCheckNormalizeError(t *testing.T) {
 		t.Fatalf("want normalize error, got %q", got)
 	}
 }
+
+// TestRunMCPCheckReportsUnreachableServer covers the case the command exists
+// for: a server that fails to start.
+//
+// RegisterTools is best-effort by design, so a connect failure is recorded on
+// the runtime and registration returns nil. runMCPCheck used to inspect only
+// that nil error and then print "is reachable", which told the user their
+// broken server was fine. The command is spawned for real here rather than
+// stubbed, because the whole failure mode was a mismatch between what
+// registration returns and what it records.
+func TestRunMCPCheckReportsUnreachableServer(t *testing.T) {
+	setTrustConfigRoot(t)
+	deps := appDeps{
+		getwd: func() (string, error) { return t.TempDir(), nil },
+		resolveMCPConfig: func(_ string, _ bool) (config.MCPConfig, error) {
+			return config.MCPConfig{Servers: map[string]config.MCPServerConfig{
+				"broken": {Type: "stdio", Command: "zero-definitely-not-a-real-binary-xyz"},
+			}}, nil
+		},
+	}
+	var out, errBuf bytes.Buffer
+	if code := runWithDeps([]string{"mcp", "check", "broken"}, &out, &errBuf, deps); code == exitSuccess {
+		t.Fatalf("mcp check must fail for a server that did not start; stdout=%q stderr=%q", out.String(), errBuf.String())
+	}
+	if got := out.String(); strings.Contains(got, "is reachable") {
+		t.Fatalf("mcp check reported a failed server as reachable: %q", got)
+	}
+	if got := errBuf.String(); !strings.Contains(got, "not reachable") {
+		t.Fatalf("want an unreachable error naming the server, got %q", got)
+	}
+}
+
+// The --json contract for an unreachable server, which the text case above does
+// not touch.
+//
+// The JSON branch is a separate return path with its own payload, so a caller
+// parsing it would otherwise be relying on a shape nothing pins: reverting the
+// check leaves it emitting status "ok" with a tool count, which is valid JSON
+// and completely wrong.
+func TestRunMCPCheckReportsUnreachableServerJSON(t *testing.T) {
+	setTrustConfigRoot(t)
+	deps := appDeps{
+		getwd: func() (string, error) { return t.TempDir(), nil },
+		resolveMCPConfig: func(_ string, _ bool) (config.MCPConfig, error) {
+			return config.MCPConfig{Servers: map[string]config.MCPServerConfig{
+				"broken": {Type: "stdio", Command: "zero-definitely-not-a-real-binary-xyz"},
+			}}, nil
+		},
+	}
+	var out, errBuf bytes.Buffer
+	if code := runWithDeps([]string{"mcp", "check", "broken", "--json"}, &out, &errBuf, deps); code == exitSuccess {
+		t.Fatalf("mcp check --json must fail for a server that did not start; stdout=%q stderr=%q", out.String(), errBuf.String())
+	}
+	var payload struct {
+		ServerName string `json:"serverName"`
+		Status     string `json:"status"`
+		Error      string `json:"error"`
+		ToolCount  int    `json:"toolCount"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &payload); err != nil {
+		t.Fatalf("decode mcp check --json output %q: %v", out.String(), err)
+	}
+	if payload.ServerName != "broken" {
+		t.Fatalf("serverName = %q, want the server that was checked", payload.ServerName)
+	}
+	if payload.Status != "unreachable" {
+		t.Fatalf("status = %q, want %q", payload.Status, "unreachable")
+	}
+	if !strings.Contains(payload.Error, "not reachable") {
+		t.Fatalf("error = %q, want it to say the server is not reachable", payload.Error)
+	}
+	// A reachable payload carries a tool count; an unreachable one must not
+	// imply zero tools were found, which reads as a server that simply exposes
+	// none.
+	if payload.ToolCount != 0 {
+		t.Fatalf("toolCount = %d, want it absent from an unreachable payload", payload.ToolCount)
+	}
+}

internal/minify/minify.go

@@ -13,8 +13,10 @@ package minify
 
 import (
 	"bytes"
+	"go/ast"
 	"go/parser"
 	"go/printer"
+	"go/scanner"
 	"go/token"
 	"path/filepath"
 	"strings"
@@ -34,8 +36,11 @@ func File(path string, content []byte) Result {
 		if out, ok := minifyGo(content); ok {
 			return Result{Content: out, Language: "go", Applied: true}
 		}
-		// Unparsable Go (a snippet, generics edge, or syntax error): fall through
-		// to the safe generic path rather than risk a partial AST reprint.
+		if out, ok := minifyGoFragment(content); ok {
+			return Result{Content: out, Language: "go-fragment", Applied: true}
+		}
+		// An incomplete fragment that cannot be wrapped safely falls through to
+		// whitespace-only normalization; no source text is guessed or discarded.
 	} else if style, ok := commentStyles[ext]; ok {
 		// Strip comments with a string-aware lexer, then collapse the whitespace the
 		// removed comments left behind. The stripper only handles languages whose
@@ -46,6 +51,133 @@ func File(path string, content []byte) Result {
 	return Result{Content: minifyGeneric(content), Language: "text", Applied: false}
 }
 
+// Fragment minifies a bounded source range without assuming lexical state from
+// text outside the range. Go remains parser-validated; other languages use the
+// conservative whitespace-only transform because a range may begin inside a
+// multiline string or block comment.
+func Fragment(path string, content []byte) Result {
+	if strings.EqualFold(filepath.Ext(path), ".go") {
+		if out, ok := minifyGo(content); ok {
+			return Result{Content: out, Language: "go", Applied: true}
+		}
+		if out, ok := minifyGoFragment(content); ok {
+			return Result{Content: out, Language: "go-fragment", Applied: true}
+		}
+	}
+	return Result{Content: minifyGeneric(content), Language: "text", Applied: false}
+}
+
+// ContextualFragment minifies a bounded fragment after checking its starting
+// position against the complete source. A Go range that begins inside a string
+// or comment must remain verbatim apart from whitespace normalization: parsing
+// that fragment by itself would mistake literal text for Go syntax.
+func ContextualFragment(path string, source, content []byte, startOffset int) Result {
+	if strings.EqualFold(filepath.Ext(path), ".go") && goFragmentStartsInsideToken(source, startOffset) {
+		return Result{Content: minifyGeneric(content), Language: "text", Applied: false}
+	}
+	return Fragment(path, content)
+}
+
+func goFragmentStartsInsideToken(source []byte, startOffset int) bool {
+	if startOffset <= 0 || startOffset >= len(source) {
+		return false
+	}
+	fset := token.NewFileSet()
+	file := fset.AddFile("", fset.Base(), len(source))
+	var lexer scanner.Scanner
+	lexer.Init(file, source, nil, scanner.ScanComments)
+	for {
+		position, kind, literal := lexer.Scan()
+		if kind == token.EOF {
+			return false
+		}
+		if kind != token.STRING && kind != token.COMMENT {
+			continue
+		}
+		start := file.Offset(position)
+		if startOffset > start && startOffset < start+len(literal) {
+			return true
+		}
+	}
+}
+
+// minifyGoFragment handles bounded reads that do not contain a package clause.
+// It finds the longest complete declaration or statement prefix that the real
+// Go parser accepts, prints that prefix without comments, and preserves any
+// incomplete tail conservatively. Ranges are intentionally capped: repeatedly
+// parsing a large broken file would otherwise turn a cheap read into O(n²) work.
+func minifyGoFragment(content []byte) (string, bool) {
+	lines := strings.Split(strings.ReplaceAll(string(content), "\r\n", "\n"), "\n")
+	if len(lines) > 1 && lines[len(lines)-1] == "" {
+		lines = lines[:len(lines)-1]
+	}
+	if len(lines) > 512 {
+		return "", false
+	}
+	for end := len(lines); end > 0; end-- {
+		prefix := strings.Join(lines[:end], "\n")
+		if compact, ok := parseGoDeclarations(prefix); ok {
+			return joinCompactPrefix(compact, lines[end:]), true
+		}
+		if compact, ok := parseGoStatements(prefix); ok {
+			return joinCompactPrefix(compact, lines[end:]), true
+		}
+	}
+	return "", false
+}
+
+func parseGoDeclarations(fragment string) (string, bool) {
+	fset := token.NewFileSet()
+	file, err := parser.ParseFile(fset, "", "package compact\n"+fragment, parser.SkipObjectResolution)
+	if err != nil || len(file.Decls) == 0 {
+		return "", false
+	}
+	nodes := make([]ast.Node, len(file.Decls))
+	for i, declaration := range file.Decls {
+		nodes[i] = declaration
+	}
+	return printGoNodes(fset, nodes)
+}
+
+func parseGoStatements(fragment string) (string, bool) {
+	fset := token.NewFileSet()
+	file, err := parser.ParseFile(fset, "", "package compact\nfunc _(){\n"+fragment+"\n}", parser.SkipObjectResolution)
+	if err != nil || len(file.Decls) != 1 {
+		return "", false
+	}
+	function, ok := file.Decls[0].(*ast.FuncDecl)
+	if !ok || function.Body == nil || len(function.Body.List) == 0 {
+		return "", false
+	}
+	nodes := make([]ast.Node, len(function.Body.List))
+	for i, statement := range function.Body.List {
+		nodes[i] = statement
+	}
+	return printGoNodes(fset, nodes)
+}
+
+func printGoNodes(fset *token.FileSet, nodes []ast.Node) (string, bool) {
+	var buf bytes.Buffer
+	cfg := printer.Config{Mode: printer.TabIndent, Tabwidth: 1}
+	for i, node := range nodes {
+		if i > 0 {
+			buf.WriteByte('\n')
+		}
+		if err := cfg.Fprint(&buf, fset, node); err != nil {
+			return "", false
+		}
+	}
+	return strings.TrimSpace(buf.String()), true
+}
+
+func joinCompactPrefix(compact string, remainder []string) string {
+	tail := minifyGeneric([]byte(strings.Join(remainder, "\n")))
+	if tail == "" {
+		return compact
+	}
+	return compact + "\n" + tail
+}
+
 // minifyGo parses Go WITHOUT comments (omitting parser.ParseComments leaves them
 // unattached) and reprints the AST, yielding valid, comment-free, gofmt-shaped
 // source with tab indentation (1 char per level). It returns ok=false on any

internal/minify/minify_test.go

@@ -39,17 +39,58 @@ func Greet(name string) string {
 	}
 }
 
-func TestMinifyGoFallsBackOnUnparsableGo(t *testing.T) {
-	// A snippet (no package clause) cannot parse as a file -> safe generic path.
+func TestMinifyGoStatementFragment(t *testing.T) {
 	r := File("snippet.go", []byte("x := 1   \n\n\n// keep me\ny := 2\n"))
-	if r.Applied {
-		t.Fatalf("expected fallback for unparsable Go, got %+v", r)
+	if !r.Applied || r.Language != "go-fragment" {
+		t.Fatalf("expected parsed Go fragment, got %+v", r)
 	}
 	if strings.Contains(r.Content, "\n\n\n") {
-		t.Errorf("generic should collapse blank runs:\n%q", r.Content)
+		t.Errorf("fragment should collapse blank runs:\n%q", r.Content)
 	}
-	if !strings.Contains(r.Content, "// keep me") {
-		t.Errorf("generic must NOT strip comments (unsafe without a parser): %q", r.Content)
+	if strings.Contains(r.Content, "keep me") || !strings.Contains(r.Content, "y := 2") {
+		t.Errorf("fragment was not compacted safely: %q", r.Content)
+	}
+}
+
+func TestMinifyGoFragmentAccepts512LinesWithFinalNewline(t *testing.T) {
+	src := strings.Repeat("x++\n", 512)
+	r := File("snippet.go", []byte(src))
+	if !r.Applied || r.Language != "go-fragment" {
+		t.Fatalf("expected 512-line fragment to remain eligible, got %+v", r)
+	}
+	if got := strings.Count(r.Content, "x++"); got != 512 {
+		t.Fatalf("compacted statement count = %d, want 512", got)
+	}
+}
+
+func TestMinifyGoDeclarationPrefixPreservesIncompleteTail(t *testing.T) {
+	src := `func Handle737(n int) (int, error) {
+	if n%2 == 1 {
+		return 0, errBad
+	}
+	return n + 737, nil
+}
+
+// Handle738 begins outside the requested range.
+func Handle738(n int) (int, error) {
+	if n%2 == 1 {`
+	r := File("range.go", []byte(src))
+	if !r.Applied || r.Language != "go-fragment" {
+		t.Fatalf("expected declaration-prefix compaction, got %+v", r)
+	}
+	if strings.Contains(r.Content, "Handle737 returns") || !strings.Contains(r.Content, "return n + 737") {
+		t.Fatalf("complete declaration was not compacted correctly:\n%s", r.Content)
+	}
+	if !strings.Contains(r.Content, "func Handle738") || !strings.Contains(r.Content, "if n%2 == 1 {") {
+		t.Fatalf("incomplete tail was not preserved:\n%s", r.Content)
+	}
+}
+
+func TestMinifyGoLargeInvalidInputFallsBackConservatively(t *testing.T) {
+	src := strings.Repeat("x := 1 // keep\n", 513)
+	r := File("broken.go", []byte(src))
+	if r.Applied || !strings.Contains(r.Content, "// keep") {
+		t.Fatalf("large invalid input must retain conservative fallback: %+v", r)
 	}
 }
 

internal/specialist/exec.go

@@ -167,10 +167,13 @@ const permissionModeUnsafe = "unsafe"
 var readOnlySpecialistTools = map[string]bool{
 	"read_file":          true,
 	"read_minified_file": true,
-	"list_directory":     true,
-	"grep":               true,
-	"glob":               true,
-	"update_plan":        true,
+	// Reads one image file through read_file's own path scoping and mutates
+	// nothing, so it carries the same auto-approval as the other pure readers.
+	"view_image":     true,
+	"list_directory": true,
+	"grep":           true,
+	"glob":           true,
+	"update_plan":    true,
 }
 
 // IsReadOnlySpecialist reports whether the named specialist resolves to a

internal/specialist/exec_test.go

@@ -255,7 +255,7 @@ func TestBuildArgsDefaultsToReadOnlyToolAllowlist(t *testing.T) {
 	if err != nil {
 		t.Fatalf("BuildArgs returned error: %v", err)
 	}
-	if !containsSequence(result.Args, []string{"--enabled-tools", "glob,grep,list_directory,read_file,read_minified_file"}) {
+	if !containsSequence(result.Args, []string{"--enabled-tools", "glob,grep,list_directory,read_file,read_minified_file,view_image"}) {
 		t.Fatalf("args missing default read-only allowlist: %#v", result.Args)
 	}
 }

internal/specialist/manifest.go

@@ -82,9 +82,9 @@ var knownMetadataKeys = map[string]bool{
 }
 
 var toolCategories = map[string][]string{
-	"read-only": {"read_file", "read_minified_file", "list_directory", "grep", "glob"},
-	"edit":      {"read_file", "read_minified_file", "list_directory", "grep", "glob", "write_file", "edit_file", "apply_patch"},
-	"execute":   {"read_file", "read_minified_file", "list_directory", "grep", "glob", "exec_command", "write_stdin", "bash"},
+	"read-only": {"read_file", "read_minified_file", "view_image", "list_directory", "grep", "glob"},
+	"edit":      {"read_file", "read_minified_file", "view_image", "list_directory", "grep", "glob", "write_file", "edit_file", "apply_patch"},
+	"execute":   {"read_file", "read_minified_file", "view_image", "list_directory", "grep", "glob", "exec_command", "write_stdin", "bash"},
 	"plan":      {"update_plan"},
 }
 
@@ -102,6 +102,7 @@ var defaultToolSelection = []string{"read-only"}
 var knownToolNames = map[string]bool{
 	"read_file":           true,
 	"read_minified_file":  true,
+	"view_image":          true,
 	"list_directory":      true,
 	"glob":                true,
 	"grep":                true,

internal/specialist/view_image_category_test.go

@@ -0,0 +1,31 @@
+package specialist
+
+import (
+	"slices"
+	"testing"
+)
+
+// view_image is a pure workspace-scoped read, so every category that already
+// grants read_file should grant it too. Categories are what specialist authors
+// actually write ("tools: [execute]"), so an omission here silently denies a
+// specialist the ability to look at a screenshot it just captured.
+func TestToolCategoriesIncludeViewImageWhereverReadFileIsGranted(t *testing.T) {
+	for category, names := range toolCategories {
+		if !slices.Contains(names, "read_file") {
+			continue
+		}
+		if !slices.Contains(names, "view_image") {
+			t.Errorf("category %q grants read_file but not view_image", category)
+		}
+	}
+}
+
+func TestResolveToolsExpandsExecuteToIncludeViewImage(t *testing.T) {
+	resolved, err := ResolveTools([]string{"execute"})
+	if err != nil {
+		t.Fatalf("ResolveTools: %v", err)
+	}
+	if !slices.Contains(resolved, "view_image") {
+		t.Errorf("execute resolved to %v, missing view_image", resolved)
+	}
+}

internal/tools/args.go

@@ -132,6 +132,34 @@ func intPtr(value int) *int {
 	return &value
 }
 
+// sandboxPermissionsDescription and additionalPermissionsDescription are shared
+// by every shell tool (bash, exec_command) so the two fields cannot drift apart.
+//
+// Both lead with "omit" on purpose. The pair used to be described only in terms
+// of what each value does, which left a model that had used
+// with_additional_permissions once (say, for `gh issue create`) no instruction
+// to stop attaching a permission object to the next, ordinary command. That
+// attachment is then rejected — correctly, but the rejection alone does not tell
+// the model that dropping the field is the fix, so it flips between an
+// unflagged permission object and a flagged empty one and never converges
+// (#845, #847). Making "omit it" the first sentence of both descriptions stops
+// the invalid shape from being emitted in the first place; the validator errors
+// are the backstop, not the teacher.
+//
+// These ride on every turn's exec_command schema, so they are kept close to the
+// length of the text they replaced; TestEagerToolSchemaTokenBudget is the ratchet.
+const (
+	sandboxPermissionsDescription = "Per-command sandbox override. Omit it unless the command needs access the " +
+		"sandbox does not already give it; use `with_additional_permissions` with a non-empty " +
+		"`additional_permissions` for sandboxed file/network access, or `require_escalated` only when the command " +
+		"must run outside the sandbox, such as host/global process, socket, service, or desktop state hidden by " +
+		"sandbox namespaces."
+
+	additionalPermissionsDescription = "Extra sandbox access. Omit this field entirely unless `sandbox_permissions` " +
+		"is `with_additional_permissions`; sending it with any other mode is rejected, as is an empty or all-false " +
+		"object. Example: {\"network\": {\"enabled\": true}}."
+)
+
 func additionalPermissionsProperties() map[string]PropertySchema {
 	return requestPermissionsProfileSchema().Properties
 }

internal/tools/bash.go

@@ -50,10 +50,10 @@ func NewScopedBashTool(workspaceRoot string, scope PathScope) Tool {
 					"command":             {Type: "string", Description: "Shell command to execute using the host shell. " + shellGuidance},
 					"cwd":                 {Type: "string", Description: "Directory to run the command in. Relative paths stay in the workspace; use an absolute path to run in a granted extra directory. Defaults to workspace root. Prefer cwd over cd when changing directories.", Default: "."},
 					"timeout_ms":          {Type: "integer", Description: "Command timeout in milliseconds.", Default: defaultBashTimeoutMS, Minimum: intPtr(1), Maximum: intPtr(maxBashTimeoutMS)},
-					"sandbox_permissions": {Type: "string", Enum: []string{string(SandboxPermissionsUseDefault), string(SandboxPermissionsWithAdditionalPermissions), string(SandboxPermissionsRequireEscalated)}, Description: "Per-command sandbox override. Defaults to `use_default`; use `with_additional_permissions` with `additional_permissions` for sandboxed file/network access, or `require_escalated` only when the command must run outside the sandbox, such as host/global process, socket, service, or desktop state hidden by sandbox namespaces.", Default: string(SandboxPermissionsUseDefault)},
+					"sandbox_permissions": {Type: "string", Enum: []string{string(SandboxPermissionsUseDefault), string(SandboxPermissionsWithAdditionalPermissions), string(SandboxPermissionsRequireEscalated)}, Description: sandboxPermissionsDescription, Default: string(SandboxPermissionsUseDefault)},
 					"additional_permissions": {
 						Type:        "object",
-						Description: "Sandboxed filesystem or network access for this command; only with `sandbox_permissions: \"with_additional_permissions\"`.",
+						Description: additionalPermissionsDescription,
 						Properties:  additionalPermissionsProperties(),
 					},
 					"justification": {Type: "string", Description: "User-facing approval question for `require_escalated`; omit otherwise."},

internal/tools/exec_command.go

@@ -91,10 +91,10 @@ func NewScopedExecCommandTool(workspaceRoot string, scope PathScope, manager *ex
 					"cwd":                 {Type: "string", Description: "Alias for workdir. Prefer workdir.", Default: "."},
 					"yield_time_ms":       {Type: "integer", Description: "Wait before yielding output. Defaults to 10000 ms; effective range is 250-30000 ms.", Default: defaultExecYieldTimeMS, Minimum: intPtr(1), Maximum: intPtr(maxExecYieldTimeMS)},
 					"max_output_tokens":   {Type: "integer", Description: "Output token budget. Defaults to 10000 tokens; larger requests may be capped by policy.", Default: defaultMaxOutputTokens, Minimum: intPtr(1), Maximum: intPtr(maxExecOutputTokenRequest)},
-					"sandbox_permissions": {Type: "string", Enum: []string{string(SandboxPermissionsUseDefault), string(SandboxPermissionsWithAdditionalPermissions), string(SandboxPermissionsRequireEscalated)}, Description: "Per-command sandbox override. Defaults to `use_default`; use `with_additional_permissions` with `additional_permissions` for sandboxed file/network access, or `require_escalated` only when the command must run outside the sandbox, such as host/global process, socket, service, or desktop state hidden by sandbox namespaces.", Default: string(SandboxPermissionsUseDefault)},
+					"sandbox_permissions": {Type: "string", Enum: []string{string(SandboxPermissionsUseDefault), string(SandboxPermissionsWithAdditionalPermissions), string(SandboxPermissionsRequireEscalated)}, Description: sandboxPermissionsDescription, Default: string(SandboxPermissionsUseDefault)},
 					"additional_permissions": {
 						Type:        "object",
-						Description: "Sandboxed filesystem or network access for this command; only with `sandbox_permissions: \"with_additional_permissions\"`.",
+						Description: additionalPermissionsDescription,
 						Properties:  additionalPermissionsProperties(),
 					},
 					"justification": {Type: "string", Description: "User-facing approval question for `require_escalated`; omit otherwise."},

internal/tools/file_safety_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"os"
 	"path/filepath"
+	"strconv"
 	"strings"
 	"testing"
 )
@@ -295,6 +296,40 @@ func TestEditFileAllowsSeenRangeAndRejectsUnseenRange(t *testing.T) {
 	}
 }
 
+func TestCanonicalLimitedReadAuthorizesEdit(t *testing.T) {
+	dir := t.TempDir()
+	var source strings.Builder
+	for line := 1; line <= 1000; line++ {
+		source.WriteString("line ")
+		source.WriteString(strconv.Itoa(line))
+		source.WriteByte('\n')
+	}
+	path := filepath.Join(dir, "f.txt")
+	if err := os.WriteFile(path, []byte(source.String()), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	registry := safetyRegistry(t, dir)
+	tracker := NewFileTracker()
+	opts := grantedOpts(tracker)
+
+	read := registry.RunWithOptions(context.Background(), "read_file", map[string]any{
+		"path": "f.txt", "offset": 500, "limit": 10,
+	}, opts)
+	if read.Status != StatusOK || read.Truncated {
+		t.Fatalf("bounded canonical read must be exact: status=%s truncated=%v meta=%#v", read.Status, read.Truncated, read.Meta)
+	}
+	if read.Meta["seen_lines"] != "500-509" {
+		t.Fatalf("bounded canonical read did not receive exact-read credit: %#v", read.Meta)
+	}
+
+	edit := registry.RunWithOptions(context.Background(), "edit_file", map[string]any{
+		"path": "f.txt", "old_string": "line 505", "new_string": "updated 505",
+	}, opts)
+	if edit.Status != StatusOK {
+		t.Fatalf("edit inside canonical read range was refused: %s", edit.Output)
+	}
+}
+
 func TestEditFileReplaceAllRecordsEveryShiftedReplacementSpan(t *testing.T) {
 	dir := t.TempDir()
 	path := filepath.Join(dir, "f.txt")

internal/tools/file_tools_test.go

@@ -38,6 +38,108 @@ func TestReadFileToolReadsLineRanges(t *testing.T) {
 	}
 }
 
+func TestReadFileToolReadsCanonicalLineRange(t *testing.T) {
+	root := t.TempDir()
+	writeTestFile(t, filepath.Join(root, "notes.txt"), "alpha\nbeta\ngamma\ndelta")
+
+	result := NewScopedReadFileTool(root, nil).Run(context.Background(), map[string]any{
+		"path":   "notes.txt",
+		"offset": 2,
+		"limit":  2,
+	})
+
+	if result.Status != StatusOK {
+		t.Fatalf("expected ok status, got %s: %s", result.Status, result.Output)
+	}
+	if result.Truncated {
+		t.Fatalf("canonical limit defines the requested range and must not mark it truncated: %#v", result.Meta)
+	}
+	if !strings.Contains(result.Output, "2 | beta") || !strings.Contains(result.Output, "3 | gamma") {
+		t.Fatalf("canonical range returned the wrong lines: %q", result.Output)
+	}
+	if strings.Contains(result.Output, "alpha") || strings.Contains(result.Output, "delta") {
+		t.Fatalf("canonical range leaked outside requested slice: %q", result.Output)
+	}
+}
+
+func TestReadFileToolCanonicalOffsetPastEnd(t *testing.T) {
+	root := t.TempDir()
+	writeTestFile(t, filepath.Join(root, "notes.txt"), "alpha\nbeta")
+
+	result := NewScopedReadFileTool(root, nil).Run(context.Background(), map[string]any{
+		"path": "notes.txt", "offset": 10,
+	})
+
+	if result.Status != StatusOK || !strings.Contains(result.Output, "offset 10 is past the end") {
+		t.Fatalf("expected canonical out-of-range message, got status=%s output=%q", result.Status, result.Output)
+	}
+	if strings.Contains(result.Output, "start_line") {
+		t.Fatalf("out-of-range message exposed legacy argument: %q", result.Output)
+	}
+}
+
+func TestReadFileToolMixedLegacyRangesPreferLines(t *testing.T) {
+	root := t.TempDir()
+	writeTestFile(t, filepath.Join(root, "notes.txt"), "alpha\nbeta\ngamma\ndelta")
+
+	result := NewScopedReadFileTool(root, nil).Run(context.Background(), map[string]any{
+		"path":        "notes.txt",
+		"start_line":  2,
+		"end_line":    3,
+		"max_lines":   2,
+		"byte_offset": 0,
+		"byte_limit":  readFileByteChunkMax,
+	})
+
+	if result.Status != StatusOK {
+		t.Fatalf("mixed legacy range should recover, got %s: %s", result.Status, result.Output)
+	}
+	if !strings.Contains(result.Output, "2 | beta") || !strings.Contains(result.Output, "3 | gamma") {
+		t.Fatalf("mixed legacy range did not prefer lines: %q", result.Output)
+	}
+}
+
+func TestReadFileToolCombinesLegacyStartWithCanonicalLimit(t *testing.T) {
+	root := t.TempDir()
+	writeTestFile(t, filepath.Join(root, "notes.txt"), "alpha\nbeta\ngamma\ndelta")
+
+	result := NewScopedReadFileTool(root, nil).Run(context.Background(), map[string]any{
+		"path": "notes.txt", "start_line": 2, "limit": 2,
+	})
+
+	if result.Status != StatusOK || result.Truncated {
+		t.Fatalf("mixed compatible range should be exact, got status=%s truncated=%v: %s", result.Status, result.Truncated, result.Output)
+	}
+	if !strings.Contains(result.Output, "2 | beta") || !strings.Contains(result.Output, "3 | gamma") {
+		t.Fatalf("mixed compatible range returned the wrong lines: %q", result.Output)
+	}
+	if strings.Contains(result.Output, "alpha") || strings.Contains(result.Output, "delta") {
+		t.Fatalf("mixed compatible range leaked outside requested slice: %q", result.Output)
+	}
+}
+
+func TestFileToolSchemasExposeOnlyCanonicalArguments(t *testing.T) {
+	readProperties := NewScopedReadFileTool(t.TempDir(), nil).Parameters().Properties
+	for _, want := range []string{"path", "offset", "limit", "byte_offset", "byte_limit"} {
+		if _, ok := readProperties[want]; !ok {
+			t.Fatalf("read_file schema missing %q", want)
+		}
+	}
+	for _, legacy := range []string{"start_line", "end_line", "max_lines"} {
+		if _, ok := readProperties[legacy]; ok {
+			t.Fatalf("read_file schema must not expose legacy argument %q", legacy)
+		}
+	}
+
+	grepProperties := NewScopedGrepTool(t.TempDir(), nil).Parameters().Properties
+	if _, ok := grepProperties["case_insensitive"]; !ok {
+		t.Fatal("grep schema missing case_insensitive")
+	}
+	if _, ok := grepProperties["-i"]; ok {
+		t.Fatal("grep schema must not expose the -i alias")
+	}
+}
+
 func TestReadFileToolMarksTruncation(t *testing.T) {
 	root := t.TempDir()
 	writeTestFile(t, filepath.Join(root, "notes.txt"), "a\nb\nc\nd\ne")
@@ -52,7 +154,7 @@ func TestReadFileToolMarksTruncation(t *testing.T) {
 	}
 	// The cut must be visible in the rendered output, not just the Truncated flag,
 	// and must tell the model where to continue.
-	if !strings.Contains(result.Output, "[truncated:") || !strings.Contains(result.Output, "start_line=3") {
+	if !strings.Contains(result.Output, "[truncated:") || !strings.Contains(result.Output, "offset=3") {
 		t.Fatalf("expected truncation marker pointing to the next line, got %q", result.Output)
 	}
 }
@@ -68,7 +170,7 @@ func TestReadFileToolAppliesByteBudget(t *testing.T) {
 	if result.Status != StatusOK || !result.Truncated {
 		t.Fatalf("expected ok+truncated, got status=%s truncated=%v", result.Status, result.Truncated)
 	}
-	if !strings.Contains(result.Output, "output exceeded") || !strings.Contains(result.Output, "start_line") {
+	if !strings.Contains(result.Output, "output exceeded") || !strings.Contains(result.Output, "offset/limit") {
 		t.Fatalf("expected byte-budget continuation hint, got %q", result.Output[len(result.Output)-200:])
 	}
 	if result.Meta["raw_bytes"] == "" || result.Meta["emitted_bytes"] == "" {

internal/tools/file_tracker.go

@@ -377,5 +377,5 @@ func fileConflictMessage(relativePath string) string {
 }
 
 func fileUnseenMessage(relativePath string) string {
-	return "Error writing " + relativePath + ": the intended change depends on content that has not been read exactly in this session. Read the affected lines with read_file, or use byte_offset/byte_limit for an oversized single line, then retry the edit."
+	return "Error writing " + relativePath + ": the intended change depends on content that has not been read exactly in this session. Read the affected lines with read_file using offset/limit (or byte_offset/byte_limit for a very long line), then retry the edit."
 }

internal/tools/grep.go

@@ -48,7 +48,6 @@ func NewScopedGrepTool(workspaceRoot string, scope PathScope) Tool {
 					"path":             {Type: "string", Description: "Directory or file to search. Relative paths stay in the workspace; use an absolute path to search a granted extra root. Defaults to workspace root.", Default: "."},
 					"glob":             {Type: "string", Description: `Optional glob filter, for example "**/*.go".`},
 					"output_mode":      {Type: "string", Description: "Output mode.", Enum: []string{"content", "files_with_matches", "count"}, Default: "content"},
-					"-i":               {Type: "boolean", Description: "Case insensitive search.", Default: false},
 					"case_insensitive": {Type: "boolean", Description: "Case insensitive search.", Default: false},
 					"head_limit":       {Type: "integer", Description: "Maximum content results to return.", Default: 50, Minimum: intPtr(1)},
 				},

internal/tools/local_capture.go

@@ -4,6 +4,8 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"github.com/Gitlawb/zero/internal/imageinput"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
 	"os"
 	"path/filepath"
 	"regexp"
@@ -345,9 +347,18 @@ func captureOKResult(driver string, request captureArtifactRequest, path string,
 	if helperOutput != "" {
 		output += "\n\n" + helperOutput
 	}
+	// Attach the artifact itself. Without this the model is told a screenshot
+	// exists and cannot look at it, which is the whole point of asking for one.
+	// A PDF or an unreadable file simply yields no image and the text stands on
+	// its own, so a capture is never failed over the attachment.
+	var images []zeroruntime.ImageBlock
+	if image, err := imageinput.LoadFile(path, ""); err == nil {
+		images = []zeroruntime.ImageBlock{image}
+	}
 	return Result{
 		Status: StatusOK,
 		Output: output,
+		Images: images,
 		Meta: map[string]string{
 			"artifact_path": path,
 			"action":        request.action,

internal/tools/read_file.go

@@ -31,16 +31,15 @@ func NewScopedReadFileTool(workspaceRoot string, scope PathScope) Tool {
 	return readFileTool{
 		baseTool: baseTool{
 			name:        "read_file",
-			description: "Read a file by 1-based inclusive line range, or by exact byte chunks for oversized single-line files.",
+			description: "Read exact file text with line numbers. Use for comments, formatting, or edits; prefer read_minified_file for initial code understanding. Use offset and limit for a line range.",
 			parameters: Schema{
 				Type: "object",
 				Properties: map[string]PropertySchema{
-					"path":        {Type: "string", Description: "Path of the file to read."},
-					"start_line":  {Type: "integer", Description: "1-based inclusive line number to start reading from.", Minimum: intPtr(1)},
-					"end_line":    {Type: "integer", Description: "1-based inclusive line number to stop reading at.", Minimum: intPtr(1)},
-					"max_lines":   {Type: "integer", Description: "Maximum number of lines to return.", Minimum: intPtr(1)},
-					"byte_offset": {Type: "integer", Description: "Zero-based byte offset for an exact chunk read. Use with byte_limit for oversized single-line files; cannot be combined with line ranges.", Minimum: intPtr(0)},
-					"byte_limit":  {Type: "integer", Description: "Maximum source bytes returned in exact byte mode. Defaults to 65536.", Default: readFileByteChunkMax, Minimum: intPtr(1), Maximum: intPtr(readFileByteChunkMax)},
+					"path":        {Type: "string", Description: "File path."},
+					"offset":      {Type: "integer", Description: "Optional 1-based source line to start from.", Minimum: intPtr(1)},
+					"limit":       {Type: "integer", Description: "Optional number of source lines to return.", Minimum: intPtr(1)},
+					"byte_offset": {Type: "integer", Description: "Optional zero-based offset for exact byte reads.", Minimum: intPtr(0)},
+					"byte_limit":  {Type: "integer", Description: "Optional byte count for an exact byte read; use with byte_offset.", Minimum: intPtr(1), Maximum: intPtr(readFileByteChunkMax)},
 				},
 				Required:             []string{"path"},
 				AdditionalProperties: false,
@@ -69,7 +68,27 @@ func (tool readFileTool) run(args map[string]any, options RunOptions, directBudg
 	if err != nil {
 		return errorResult("Error: Invalid arguments for read_file: " + err.Error())
 	}
-	startLine, err := intArg(args, "start_line", 1, 1, 0)
+	_, hasOffset := args["offset"]
+	_, hasLimit := args["limit"]
+	_, hasStartLine := args["start_line"]
+	_, hasEndLine := args["end_line"]
+	_, hasMaxLines := args["max_lines"]
+	if hasOffset && hasStartLine {
+		return errorResult("Error: Invalid arguments for read_file: use either offset or start_line, not both")
+	}
+	if hasLimit && (hasEndLine || hasMaxLines) {
+		return errorResult("Error: Invalid arguments for read_file: use limit instead of end_line or max_lines, not both")
+	}
+
+	startKey := "offset"
+	if !hasOffset && hasStartLine {
+		startKey = "start_line"
+	}
+	startLine, err := intArg(args, startKey, 1, 1, 0)
+	if err != nil {
+		return errorResult("Error: Invalid arguments for read_file: " + err.Error())
+	}
+	limit, err := intArg(args, "limit", 0, 1, 0)
 	if err != nil {
 		return errorResult("Error: Invalid arguments for read_file: " + err.Error())
 	}
@@ -81,19 +100,25 @@ func (tool readFileTool) run(args map[string]any, options RunOptions, directBudg
 	if err != nil {
 		return errorResult("Error: Invalid arguments for read_file: " + err.Error())
 	}
+	if hasLimit {
+		const maxInt = int(^uint(0) >> 1)
+		endLine = startLine + min(limit-1, maxInt-startLine)
+	}
+	hasLineRange := hasOffset || hasLimit || hasStartLine || hasEndLine || hasMaxLines
 	_, hasByteOffset := args["byte_offset"]
 	_, hasByteLimit := args["byte_limit"]
-	byteMode := hasByteOffset || hasByteLimit
-	byteOffset, err := intArg(args, "byte_offset", 0, 0, 0)
-	if err != nil {
-		return errorResult("Error: Invalid arguments for read_file: " + err.Error())
-	}
-	byteLimit, err := intArg(args, "byte_limit", readFileByteChunkMax, 1, readFileByteChunkMax)
-	if err != nil {
-		return errorResult("Error: Invalid arguments for read_file: " + err.Error())
-	}
-	if byteMode && (args["start_line"] != nil || args["end_line"] != nil || args["max_lines"] != nil) {
-		return errorResult("Error: Invalid arguments for read_file: byte_offset/byte_limit cannot be combined with line ranges")
+	byteMode := (hasByteOffset || hasByteLimit) && !hasLineRange
+	byteOffset := 0
+	byteLimit := readFileByteChunkMax
+	if byteMode {
+		byteOffset, err = intArg(args, "byte_offset", 0, 0, 0)
+		if err != nil {
+			return errorResult("Error: Invalid arguments for read_file: " + err.Error())
+		}
+		byteLimit, err = intArg(args, "byte_limit", readFileByteChunkMax, 1, readFileByteChunkMax)
+		if err != nil {
+			return errorResult("Error: Invalid arguments for read_file: " + err.Error())
+		}
 	}
 
 	absolutePath, relativePath, err := resolveScopedReadPath(tool.workspaceRoot, tool.scope, requestedPath)
@@ -169,7 +194,7 @@ func renderedReadRange(total, start, end, maxLines int) (int, int) {
 
 func renderReadFileRange(absolutePath string, relativePath string, total int, startLine int, endLine int, maxLines int, maxBytes int) Result {
 	if startLine > total {
-		return okResult(fmt.Sprintf("File: %s\n(start_line %d is past the end of the file, which has %d lines)", relativePath, startLine, total))
+		return okResult(fmt.Sprintf("File: %s\n(offset %d is past the end of the file, which has %d lines)", relativePath, startLine, total))
 	}
 	if endLine == 0 || endLine > total {
 		endLine = total
@@ -208,9 +233,9 @@ func renderReadFileRange(absolutePath string, relativePath string, total int, st
 	// Tool.Run calls preserve the legacy prefix-only byte budget.
 	var budgetedOutput *outputBudgetBuilder
 	if maxBytes > 0 {
-		budgetedOutput = newOutputBudgetBuilder(maxBytes, "use start_line/end_line or max_lines for normal files; use byte_offset/byte_limit for an oversized single line")
+		budgetedOutput = newOutputBudgetBuilder(maxBytes, "use offset/limit to request a smaller line range")
 	} else {
-		budgetedOutput = newHeadTailOutputBudgetBuilder(readOutputBudgetBytes, "use start_line/end_line or max_lines for normal files; use byte_offset/byte_limit for an oversized single line")
+		budgetedOutput = newHeadTailOutputBudgetBuilder(readOutputBudgetBytes, "use offset/limit to request a smaller line range")
 	}
 	budgetedOutput.WriteString(header)
 	budgetedOutput.WriteString("\n")
@@ -224,9 +249,9 @@ func renderReadFileRange(absolutePath string, relativePath string, total int, st
 	}
 	if truncated {
 		// The Truncated flag alone is invisible to the model in the rendered
-		// output, so it cannot tell a max_lines cut from a complete read. Make the
+		// output, so it cannot tell a limit cut from a complete read. Make the
 		// cut explicit and tell it how to continue.
-		budgetedOutput.WriteString(fmt.Sprintf("\n\n[truncated: %d more line(s) in the requested range not shown; set start_line=%d to continue]", endLine-lastLine, lastLine+1))
+		budgetedOutput.WriteString(fmt.Sprintf("\n\n[truncated: %d more line(s) in the requested range not shown; set offset=%d to continue]", endLine-lastLine, lastLine+1))
 	}
 
 	budgeted := budgetedOutput.Result()
@@ -241,7 +266,7 @@ func renderReadFileRange(absolutePath string, relativePath string, total int, st
 		if budgeted.Truncated {
 			meta["truncation_reason"] = "byte_budget"
 		} else {
-			meta["truncation_reason"] = "max_lines"
+			meta["truncation_reason"] = "limit"
 		}
 	}
 	return Result{

internal/tools/read_minified_file.go

@@ -28,12 +28,13 @@ func NewScopedReadMinifiedFileTool(workspaceRoot string, scope PathScope) Tool {
 	return readMinifiedFileTool{
 		baseTool: baseTool{
 			name:        "read_minified_file",
-			description: "Read a source file in a dense, token-cheap form: comments and redundant whitespace removed, no line numbers. Use it to scan or understand code for far fewer tokens than read_file. For exact text, comments, line numbers, or before editing, use read_file instead.",
-			deferred:    true,
+			description: "Read source code in a token-efficient, language-aware form. Prefer this for initial understanding; use read_file for exact text or line numbers.",
 			parameters: Schema{
 				Type: "object",
 				Properties: map[string]PropertySchema{
-					"path": {Type: "string", Description: "Path of the file to read in minified form."},
+					"path":   {Type: "string", Description: "Source file path."},
+					"offset": {Type: "integer", Description: "Optional 1-based source line to start from.", Minimum: intPtr(1)},
+					"limit":  {Type: "integer", Description: "Optional maximum number of source lines to minify.", Minimum: intPtr(1)},
 				},
 				Required:             []string{"path"},
 				AdditionalProperties: false,
@@ -59,6 +60,14 @@ func (tool readMinifiedFileTool) run(args map[string]any, options RunOptions, di
 	if err != nil {
 		return errorResult("Error: Invalid arguments for read_minified_file: " + err.Error())
 	}
+	offset, err := intArg(args, "offset", 1, 1, 0)
+	if err != nil {
+		return errorResult("Error: Invalid arguments for read_minified_file: " + err.Error())
+	}
+	limit, err := intArg(args, "limit", 0, 1, 0)
+	if err != nil {
+		return errorResult("Error: Invalid arguments for read_minified_file: " + err.Error())
+	}
 
 	absolutePath, relativePath, err := resolveScopedReadPath(tool.workspaceRoot, tool.scope, requestedPath)
 	if err != nil {
@@ -75,11 +84,19 @@ func (tool readMinifiedFileTool) run(args map[string]any, options RunOptions, di
 	info, _ := os.Stat(absolutePath)
 	options.FileTracker.Record(absolutePath, content, info)
 
-	result := minify.File(relativePath, content)
-	rawLines := lineCount(string(content))
+	selected := selectSourceLines(content, offset, limit)
+	if selected.pastEnd {
+		return okResult(fmt.Sprintf("File: %s\n(offset %d is past the end of the file, which has %d lines)", relativePath, offset, selected.totalLines))
+	}
+	ranged := offset > 1 || limit > 0
+	result := minify.File(relativePath, selected.content)
+	if ranged {
+		result = minify.ContextualFragment(relativePath, content, selected.content, selected.startByte)
+	}
+	rawLines := lineCount(string(selected.content))
 	minLines := lineCount(result.Content)
 	pct := 0
-	if rawBytes := len(content); rawBytes > 0 {
+	if rawBytes := len(selected.content); rawBytes > 0 {
 		if saved := rawBytes - len(result.Content); saved > 0 {
 			pct = saved * 100 / rawBytes
 		}
@@ -89,12 +106,15 @@ func (tool readMinifiedFileTool) run(args map[string]any, options RunOptions, di
 	if result.Applied {
 		header = fmt.Sprintf("File: %s — minified %s view (comments stripped, no line numbers; %d→%d lines, ~%d%% fewer bytes). For exact text/comments or before editing, use read_file.",
 			relativePath, result.Language, rawLines, minLines, pct)
+	} else if ranged {
+		header = fmt.Sprintf("File: %s — safe ranged view (whitespace normalized, no line numbers; %d→%d lines; context-sensitive stripping disabled because the range may begin inside a multiline construct). For exact text, use read_file.",
+			relativePath, rawLines, minLines)
 	} else {
 		header = fmt.Sprintf("File: %s — whitespace-normalized view (no line numbers; %d→%d lines; full minification not available for this file type). For exact text, use read_file.",
 			relativePath, rawLines, minLines)
 	}
 
-	rawBytes := len(content)
+	rawBytes := len(selected.content)
 	compactBytes := len(result.Content)
 	savedTokens := 0
 	if savedBytes := rawBytes - compactBytes; savedBytes > 0 {
@@ -113,11 +133,54 @@ func (tool readMinifiedFileTool) run(args map[string]any, options RunOptions, di
 	meta["estimated_tokens_saved"] = strconv.Itoa(savedTokens)
 	toolResult := Result{Status: StatusOK, Output: output, Meta: meta}
 	if directBudget {
-		return applyLegacyByteBudgetToResult(toolResult, readOutputBudgetBytes, "use read_file with start_line/end_line or max_lines for normal files, or byte_offset/byte_limit for an oversized single line")
+		return applyLegacyByteBudgetToResult(toolResult, readOutputBudgetBytes, "use offset/limit to request a smaller source range, or read_file when exact text is required")
 	}
 	return toolResult
 }
 
+type sourceSelection struct {
+	content    []byte
+	startByte  int
+	totalLines int
+	pastEnd    bool
+}
+
+func selectSourceLines(content []byte, offset, limit int) sourceSelection {
+	if offset <= 1 && limit == 0 {
+		return sourceSelection{content: content, totalLines: sourceLineCount(content)}
+	}
+	lines := strings.Split(string(content), "\n")
+	totalLines := sourceLineCount(content)
+	start := offset - 1
+	if start >= totalLines {
+		return sourceSelection{startByte: len(content), totalLines: totalLines, pastEnd: true}
+	}
+	startByte := 0
+	for index := 0; index < start; index++ {
+		startByte += len(lines[index]) + 1
+	}
+	end := totalLines
+	if limit > 0 && limit < end-start {
+		end = start + limit
+	}
+	return sourceSelection{
+		content:    []byte(strings.Join(lines[start:end], "\n")),
+		startByte:  startByte,
+		totalLines: totalLines,
+	}
+}
+
+func sourceLineCount(content []byte) int {
+	if len(content) == 0 {
+		return 1
+	}
+	lines := strings.Count(string(content), "\n")
+	if content[len(content)-1] != '\n' {
+		lines++
+	}
+	return lines
+}
+
 // lineCount reports the number of newline-separated lines in s (an empty string
 // counts as 0 lines, matching how a reader perceives an empty file).
 func lineCount(s string) int {

internal/tools/read_minified_file_test.go

@@ -45,6 +45,123 @@ func TestReadMinifiedFileRejectsTraversal(t *testing.T) {
 	}
 }
 
+func TestReadMinifiedFileSelectsSourceLineRangeBeforeMinifying(t *testing.T) {
+	dir := t.TempDir()
+	src := "package demo\n\nfunc One() int { return 1 }\n\nfunc Two() int { return 2 }\n"
+	if err := os.WriteFile(filepath.Join(dir, "f.go"), []byte(src), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	res := NewScopedReadMinifiedFileTool(dir, nil).Run(context.Background(), map[string]any{
+		"path": "f.go", "offset": 5, "limit": 1,
+	})
+	if res.Status != StatusOK || !strings.Contains(res.Output, "func Two") || strings.Contains(res.Output, "func One") {
+		t.Fatalf("unexpected ranged compact read: status=%s\n%s", res.Status, res.Output)
+	}
+}
+
+func TestReadMinifiedFileRangesPreserveUnknownLexicalContext(t *testing.T) {
+	tests := []struct {
+		name string
+		path string
+		src  string
+		want []string
+	}{
+		{
+			name: "multiline string",
+			path: "snippet.py",
+			src:  "value = \"\"\"\n# literal text\nstill literal\n\"\"\"\nprint(value)\n",
+			want: []string{"# literal text", "still literal"},
+		},
+		{
+			name: "template literal",
+			path: "snippet.js",
+			src:  "const value = `\n// literal text\nstill literal\n`;\n",
+			want: []string{"// literal text", "still literal"},
+		},
+		{
+			name: "block comment",
+			path: "snippet.c",
+			src:  "/* open\ncomment body\n*/\nint live;\n",
+			want: []string{"comment body", "*/"},
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			dir := t.TempDir()
+			if err := os.WriteFile(filepath.Join(dir, tc.path), []byte(tc.src), 0o644); err != nil {
+				t.Fatal(err)
+			}
+			res := NewScopedReadMinifiedFileTool(dir, nil).Run(context.Background(), map[string]any{
+				"path": tc.path, "offset": 2, "limit": 2,
+			})
+			if res.Status != StatusOK || res.Meta["compacted"] != "false" {
+				t.Fatalf("ranged read must use conservative normalization: status=%s meta=%#v\n%s", res.Status, res.Meta, res.Output)
+			}
+			for _, want := range tc.want {
+				if !strings.Contains(res.Output, want) {
+					t.Fatalf("ranged read lost lexical content %q:\n%s", want, res.Output)
+				}
+			}
+		})
+	}
+}
+
+func TestReadMinifiedFileRangeInsideGoRawStringIsConservative(t *testing.T) {
+	dir := t.TempDir()
+	src := "package demo\nvar value = `first\nSECRET-MARKER-ONE\n// literal line\nlast`\n"
+	if err := os.WriteFile(filepath.Join(dir, "f.go"), []byte(src), 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	res := NewScopedReadMinifiedFileTool(dir, nil).Run(context.Background(), map[string]any{
+		"path": "f.go", "offset": 3, "limit": 2,
+	})
+	if res.Status != StatusOK || res.Meta["compacted"] != "false" {
+		t.Fatalf("raw-string range must use conservative normalization: status=%s meta=%#v\n%s", res.Status, res.Meta, res.Output)
+	}
+	for _, want := range []string{"SECRET-MARKER-ONE", "// literal line"} {
+		if !strings.Contains(res.Output, want) {
+			t.Fatalf("raw-string range lost or rewrote %q:\n%s", want, res.Output)
+		}
+	}
+	if strings.Contains(res.Output, "SECRET - MARKER - ONE") {
+		t.Fatalf("raw-string contents were parsed as Go code:\n%s", res.Output)
+	}
+}
+
+func TestReadMinifiedFileCanonicalOffsetPastEnd(t *testing.T) {
+	dir := t.TempDir()
+	if err := os.WriteFile(filepath.Join(dir, "f.go"), []byte("package demo\nvar value = 1\n"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	res := NewScopedReadMinifiedFileTool(dir, nil).Run(context.Background(), map[string]any{
+		"path": "f.go", "offset": 10,
+	})
+	if res.Status != StatusOK || !strings.Contains(res.Output, "offset 10 is past the end") {
+		t.Fatalf("expected canonical out-of-range response, got status=%s output=%q", res.Status, res.Output)
+	}
+}
+
+func TestReadMinifiedFileMaximumLimitDoesNotOverflow(t *testing.T) {
+	dir := t.TempDir()
+	if err := os.WriteFile(filepath.Join(dir, "notes.txt"), []byte("alpha\nbeta\ngamma\n"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	const maxInt = int(^uint(0) >> 1)
+
+	res := NewScopedReadMinifiedFileTool(dir, nil).Run(context.Background(), map[string]any{
+		"path": "notes.txt", "offset": 2, "limit": maxInt,
+	})
+	if res.Status != StatusOK {
+		t.Fatalf("maximum limit should return the remaining range without panicking: status=%s output=%q", res.Status, res.Output)
+	}
+	if !strings.Contains(res.Output, "beta") || !strings.Contains(res.Output, "gamma") || strings.Contains(res.Output, "alpha") {
+		t.Fatalf("maximum limit returned the wrong range: %q", res.Output)
+	}
+}
+
 func TestReadMinifiedFileAppliesByteBudget(t *testing.T) {
 	dir := t.TempDir()
 	if err := os.WriteFile(filepath.Join(dir, "large.txt"), []byte(strings.Repeat("0123456789abcdef\n", 9000)), 0o644); err != nil {

internal/tools/registry.go

@@ -297,8 +297,11 @@ func scrubResultSecrets(res Result) Result {
 
 func CoreReadOnlyToolsScoped(workspaceRoot string, scope PathScope) []Tool {
 	return []Tool{
-		NewScopedReadFileTool(workspaceRoot, scope),
 		NewScopedReadMinifiedFileTool(workspaceRoot, scope),
+		NewScopedReadFileTool(workspaceRoot, scope),
+		// view_image is read-only and shares read_file's path scoping, so it
+		// belongs with the other readers rather than behind a gate.
+		NewScopedViewImageTool(workspaceRoot, scope),
 		NewScopedListDirectoryTool(workspaceRoot, scope),
 		NewScopedGlobTool(workspaceRoot, scope),
 		NewScopedGrepTool(workspaceRoot, scope),

internal/tools/registry_test.go

@@ -30,8 +30,8 @@ func tempDirOutsideDefaultTemp(t *testing.T) string {
 
 func TestCoreReadOnlyToolsExposeSafeMetadata(t *testing.T) {
 	toolset := CoreReadOnlyToolsScoped(t.TempDir(), nil)
-	if len(toolset) != 9 {
-		t.Fatalf("expected 9 core read-only tools, got %d", len(toolset))
+	if len(toolset) != 10 {
+		t.Fatalf("expected 10 core read-only tools, got %d", len(toolset))
 	}
 
 	seen := map[string]bool{}
@@ -71,6 +71,11 @@ func TestCoreReadOnlyToolsExposeSafeMetadata(t *testing.T) {
 	if !seen[RequestPermissionsToolName] {
 		t.Fatalf("expected %s in core read-only tools", RequestPermissionsToolName)
 	}
+	// By name as well as by count: the count alone still passes if one reader is
+	// swapped for another.
+	if !seen[ViewImageToolName] {
+		t.Fatalf("expected %s in core read-only tools", ViewImageToolName)
+	}
 }
 
 func TestRegistryAllReturnsToolsSortedByName(t *testing.T) {
@@ -518,25 +523,24 @@ func TestIsDeferralEligibleDecouplesFromDeferred(t *testing.T) {
 
 func TestOptionalBuiltinsUseDeferredDiscovery(t *testing.T) {
 	wantDeferred := map[string]bool{
-		"bash":               true,
-		"browser_action":     true,
-		"browser_click":      true,
-		"browser_connect":    true,
-		"browser_install":    true,
-		"browser_launch":     true,
-		"browser_open":       true,
-		"browser_press":      true,
-		"browser_snapshot":   true,
-		"browser_type":       true,
-		"capture_artifact":   true,
-		"desktop_action":     true,
-		"desktop_snapshot":   true,
-		"desktop_windows":    true,
-		"lsp_navigate":       true,
-		"read_minified_file": true,
-		"terminal_session":   true,
-		"web_fetch":          true,
-		"web_search":         true,
+		"bash":             true,
+		"browser_action":   true,
+		"browser_click":    true,
+		"browser_connect":  true,
+		"browser_install":  true,
+		"browser_launch":   true,
+		"browser_open":     true,
+		"browser_press":    true,
+		"browser_snapshot": true,
+		"browser_type":     true,
+		"capture_artifact": true,
+		"desktop_action":   true,
+		"desktop_snapshot": true,
+		"desktop_windows":  true,
+		"lsp_navigate":     true,
+		"terminal_session": true,
+		"web_fetch":        true,
+		"web_search":       true,
 	}
 	for _, tool := range BuiltinCatalog(t.TempDir()) {
 		if _, listed := wantDeferred[tool.Name()]; listed {
@@ -544,7 +548,7 @@ func TestOptionalBuiltinsUseDeferredDiscovery(t *testing.T) {
 				t.Errorf("%s should be deferred", tool.Name())
 			}
 			delete(wantDeferred, tool.Name())
-		} else if tool.Name() == "read_file" || tool.Name() == "grep" || tool.Name() == ExecCommandToolName || tool.Name() == "edit_file" {
+		} else if tool.Name() == "read_file" || tool.Name() == "read_minified_file" || tool.Name() == "grep" || tool.Name() == ExecCommandToolName || tool.Name() == "edit_file" {
 			if IsDeferred(tool) {
 				t.Errorf("essential tool %s must stay eager", tool.Name())
 			}

internal/tools/sandbox_permissions_schema_test.go

@@ -0,0 +1,72 @@
+package tools
+
+import (
+	"strings"
+	"testing"
+)
+
+// The cheapest fix for #845/#847 is that the invalid shape is never emitted.
+// A model only learns "omit it" from the schema it is shown, so both shell
+// tools must say so in the field descriptions themselves — the validator errors
+// are the backstop, and by the time they fire a turn has already been burned.
+func TestShellToolsAdvertiseOmittingSandboxPermissionFields(t *testing.T) {
+	root := t.TempDir()
+	for _, shell := range []struct {
+		name string
+		tool Tool
+	}{
+		{"bash", NewScopedBashTool(root, nil)},
+		{"exec_command", NewScopedExecCommandTool(root, nil, nil)},
+	} {
+		properties := shell.tool.Parameters().Properties
+
+		mode, ok := properties["sandbox_permissions"]
+		if !ok {
+			t.Fatalf("%s: sandbox_permissions property missing", shell.name)
+		}
+		if !strings.Contains(mode.Description, "Omit it") {
+			t.Errorf("%s: sandbox_permissions description = %q, want it to lead with omitting the field", shell.name, mode.Description)
+		}
+		// Deliberately NOT "read-only". An earlier revision said every read-only
+		// command could omit the override, and that is false: `curl` and `gh` are
+		// read-only and still need sandboxed network. Conditioning omission on
+		// whether the command needs access it does not already have is the claim
+		// that is actually true, and the one a model can act on without first
+		// having to decide what counts as read-only.
+		if !strings.Contains(mode.Description, "unless the command needs access") {
+			t.Errorf("%s: sandbox_permissions description = %q, want omission conditioned on needing extra access", shell.name, mode.Description)
+		}
+
+		extra, ok := properties["additional_permissions"]
+		if !ok {
+			t.Fatalf("%s: additional_permissions property missing", shell.name)
+		}
+		if !strings.Contains(extra.Description, "Omit this field entirely") {
+			t.Errorf("%s: additional_permissions description = %q, want an explicit instruction to omit the field", shell.name, extra.Description)
+		}
+		// Both invalid shapes from the issue must be named as invalid up front,
+		// so neither is emitted "to be safe".
+		if !strings.Contains(extra.Description, "with any other mode is rejected") {
+			t.Errorf("%s: additional_permissions description = %q, want it to say the unflagged shape is rejected", shell.name, extra.Description)
+		}
+		if !strings.Contains(extra.Description, "empty or all-false object") {
+			t.Errorf("%s: additional_permissions description = %q, want it to say an empty payload is rejected", shell.name, extra.Description)
+		}
+	}
+}
+
+// The two shell tools describe the same two fields; they were copy-pasted and
+// must not drift apart again, or a model would be taught different rules by
+// bash and exec_command in the same session.
+func TestShellToolsShareSandboxPermissionDescriptions(t *testing.T) {
+	root := t.TempDir()
+	bash := NewScopedBashTool(root, nil).Parameters().Properties
+	exec := NewScopedExecCommandTool(root, nil, nil).Parameters().Properties
+
+	for _, field := range []string{"sandbox_permissions", "additional_permissions"} {
+		if bash[field].Description != exec[field].Description {
+			t.Errorf("%s description differs between bash and exec_command:\n bash = %q\n exec = %q",
+				field, bash[field].Description, exec[field].Description)
+		}
+	}
+}

internal/tools/shell_output_reducer.go

@@ -8,6 +8,12 @@ import (
 
 const commandReducerMinPassingLines = 12
 
+type commandOutputReducer struct {
+	kind   string
+	label  string
+	reduce func(string) (string, int)
+}
+
 // reduceCommandOutput removes repetitive success-only lines from recognized
 // simple commands. Unsupported or compound shell input passes through exactly,
 // and every changed result retains the pre-reduction output as an artifact.
@@ -16,11 +22,15 @@ func reduceCommandOutput(toolName string, args map[string]any, result Result) Re
 		return result
 	}
 	command, err := commandTextForReducer(toolName, args)
-	if err != nil || !isSimpleGoTestCommand(command) {
+	if err != nil {
+		return result
+	}
+	reducer, ok := commandOutputReducerFor(command)
+	if !ok {
 		return result
 	}
 
-	reduced, omitted := reduceGoTestPassingPackages(result.Output)
+	reduced, omitted := reducer.reduce(result.Output)
 	if omitted < commandReducerMinPassingLines || len(reduced) >= len(result.Output) {
 		return result
 	}
@@ -31,13 +41,14 @@ func reduceCommandOutput(toolName string, args map[string]any, result Result) Re
 		return result
 	}
 
-	notice := fmt.Sprintf("[zero] compacted %d passing package lines; exact output saved to %s", omitted, spillPath)
+	notice := fmt.Sprintf("[zero] compacted %d %s; exact output saved to %s", omitted, reducer.label, spillPath)
 	result.Output = strings.TrimRight(reduced, "\n") + "\n" + notice
 	result.Truncated = true
 	if result.Meta == nil {
 		result.Meta = map[string]string{}
 	}
 	result.Meta["command_output_reduced"] = "true"
+	result.Meta["command_output_reducer"] = reducer.kind
 	result.Meta["command_output_omitted_lines"] = strconv.Itoa(omitted)
 	result.Meta["command_output_original_bytes"] = strconv.Itoa(originalBytes)
 	result.Meta["command_output_original_tokens"] = strconv.Itoa(originalTokens)
@@ -54,23 +65,207 @@ func commandTextForReducer(toolName string, args map[string]any) (string, error)
 	return execCommandArg(args)
 }
 
-func isSimpleGoTestCommand(command string) bool {
+func simpleCommandFields(command string) ([]string, bool) {
 	command = strings.TrimSpace(command)
+	if separator := strings.LastIndexAny(command, " \t"); separator >= 0 && command[separator+1:] == "2>&1" {
+		command = strings.TrimSpace(command[:separator])
+	}
 	if command == "" || strings.ContainsAny(command, "|;&<>\n\r`") || strings.Contains(command, "$(") {
-		return false
+		return nil, false
 	}
 	fields := strings.Fields(command)
-	return len(fields) >= 2 && fields[0] == "go" && fields[1] == "test"
+	if len(fields) == 0 {
+		return nil, false
+	}
+	return fields, true
+}
+
+func commandOutputReducerFor(command string) (commandOutputReducer, bool) {
+	fields, ok := simpleCommandFields(command)
+	if !ok {
+		return commandOutputReducer{}, false
+	}
+	if len(fields) >= 2 && fields[0] == "go" && fields[1] == "test" {
+		return commandOutputReducer{kind: "go_test", label: "passing package lines", reduce: reduceGoTestPassingPackages}, true
+	}
+	if len(fields) >= 2 && fields[0] == "cargo" {
+		switch fields[1] {
+		case "test":
+			return commandOutputReducer{kind: "cargo_test", label: "passing Rust test lines", reduce: reduceCargoTestPassingLines}, true
+		case "build", "check":
+			return commandOutputReducer{kind: "cargo_build", label: "successful Rust build progress lines", reduce: reduceCargoBuildProgressLines}, true
+		}
+	}
+	if isPytestCommand(fields) {
+		return commandOutputReducer{kind: "pytest", label: "passing pytest progress lines", reduce: reducePytestPassingLines}, true
+	}
+	if isJavaScriptTestCommand(fields) {
+		return commandOutputReducer{kind: "javascript_test", label: "passing JavaScript test-file lines", reduce: reduceJavaScriptTestPassingLines}, true
+	}
+	return commandOutputReducer{}, false
+}
+
+func isPytestCommand(fields []string) bool {
+	if fields[0] == "pytest" || fields[0] == "py.test" {
+		return true
+	}
+	return len(fields) >= 3 && (fields[0] == "python" || fields[0] == "python3") && fields[1] == "-m" && fields[2] == "pytest"
+}
+
+func isJavaScriptTestCommand(fields []string) bool {
+	switch fields[0] {
+	case "jest", "vitest":
+		return true
+	case "npx":
+		return len(fields) >= 2 && (fields[1] == "jest" || fields[1] == "vitest")
+	case "npm", "pnpm", "yarn":
+		if len(fields) >= 2 && fields[1] == "test" {
+			return true
+		}
+		if len(fields) >= 3 && fields[1] == "run" && fields[2] == "test" {
+			return true
+		}
+		return len(fields) >= 3 && fields[1] == "exec" && (fields[2] == "jest" || fields[2] == "vitest")
+	default:
+		return false
+	}
 }
 
 func reduceGoTestPassingPackages(output string) (string, int) {
+	return reduceRepeatedOutputLines(output, "passing package lines", func(line string) bool {
+		packageSuccess := strings.HasPrefix(line, "ok  \t") || strings.HasPrefix(line, "?   \t") ||
+			strings.HasPrefix(line, "ok  ") || strings.HasPrefix(line, "?   ")
+		return packageSuccess && !strings.Contains(line, "coverage:") && !strings.Contains(line, "[no test files]")
+	})
+}
+
+func reduceCargoTestPassingLines(output string) (string, int) {
+	if !strings.Contains(output, "test result:") {
+		return output, 0
+	}
+	return reduceRepeatedOutputLines(output, "passing Rust test lines", func(line string) bool {
+		trimmed := strings.TrimSpace(line)
+		return strings.HasPrefix(trimmed, "test ") && strings.HasSuffix(trimmed, " ... ok")
+	})
+}
+
+func reduceCargoBuildProgressLines(output string) (string, int) {
+	if !outputHasCargoFinishedSummary(output) {
+		return output, 0
+	}
+	return reduceRepeatedOutputLines(output, "successful Rust build progress lines", func(line string) bool {
+		trimmed := strings.TrimSpace(line)
+		return strings.HasPrefix(trimmed, "Checking ") || strings.HasPrefix(trimmed, "Compiling ")
+	})
+}
+
+func reducePytestPassingLines(output string) (string, int) {
+	if !outputHasPytestSummary(output) {
+		return output, 0
+	}
+	return reduceRepeatedOutputLines(output, "passing pytest progress lines", isPassingPytestProgressLine)
+}
+
+func isPassingPytestProgressLine(line string) bool {
+	trimmed := strings.TrimSpace(line)
+	open := strings.LastIndexByte(trimmed, '[')
+	if open < 0 || !strings.HasSuffix(trimmed, "%]") {
+		return false
+	}
+	progress := strings.TrimSpace(trimmed[:open])
+	fields := strings.Fields(progress)
+	if len(fields) == 0 {
+		return false
+	}
+	marks := fields[len(fields)-1]
+	return marks != "" && strings.Trim(marks, ".") == ""
+}
+
+func reduceJavaScriptTestPassingLines(output string) (string, int) {
+	if !outputHasJavaScriptTestSummary(output) {
+		return output, 0
+	}
+	return reduceRepeatedOutputLines(output, "passing JavaScript test-file lines", func(line string) bool {
+		trimmed := strings.TrimSpace(line)
+		return strings.HasPrefix(line, "PASS ") || strings.HasPrefix(trimmed, "✓ ") || strings.HasPrefix(trimmed, "✔ ")
+	})
+}
+
+func outputHasCargoFinishedSummary(output string) bool {
+	for _, line := range strings.Split(output, "\n") {
+		trimmed := strings.TrimSpace(line)
+		if strings.HasPrefix(trimmed, "Finished `") && strings.Contains(trimmed, "` profile ") &&
+			strings.Contains(trimmed, " target(s) in ") {
+			return true
+		}
+	}
+	return false
+}
+
+func outputHasPytestSummary(output string) bool {
+	for _, line := range strings.Split(output, "\n") {
+		summary := strings.Trim(strings.TrimSpace(line), "= ")
+		if summary == "short test summary info" || isPytestResultSummary(summary) {
+			return true
+		}
+	}
+	return false
+}
+
+func isPytestResultSummary(summary string) bool {
+	fields := strings.Fields(summary)
+	if len(fields) < 4 {
+		return false
+	}
+	first := strings.Trim(fields[0], ",;:()")
+	if _, err := strconv.Atoi(first); err != nil || !summaryHasCountedStatus(summary, "passed") {
+		return false
+	}
+	for index := 1; index+1 < len(fields); index++ {
+		if fields[index] != "in" {
+			continue
+		}
+		duration := strings.TrimSuffix(strings.Trim(fields[index+1], ",;:()"), "s")
+		if _, err := strconv.ParseFloat(duration, 64); err == nil {
+			return true
+		}
+	}
+	return false
+}
+
+func outputHasJavaScriptTestSummary(output string) bool {
+	for _, line := range strings.Split(output, "\n") {
+		trimmed := strings.TrimSpace(line)
+		if strings.HasPrefix(trimmed, "Test Suites:") && summaryHasCountedStatus(trimmed, "passed") {
+			return true
+		}
+		if strings.HasPrefix(trimmed, "Test Files ") && summaryHasCountedStatus(trimmed, "passed") {
+			return true
+		}
+	}
+	return false
+}
+
+func summaryHasCountedStatus(summary string, status string) bool {
+	fields := strings.Fields(summary)
+	for index := 1; index < len(fields); index++ {
+		if strings.Trim(fields[index], ",;:()") != status {
+			continue
+		}
+		count := strings.Trim(fields[index-1], ",;:()")
+		if _, err := strconv.Atoi(count); err == nil {
+			return true
+		}
+	}
+	return false
+}
+
+func reduceRepeatedOutputLines(output string, label string, omit func(string) bool) (string, int) {
 	lines := strings.Split(output, "\n")
 	kept := make([]string, 0, len(lines))
 	omitted := 0
 	for _, line := range lines {
-		packageSuccess := strings.HasPrefix(line, "ok  \t") || strings.HasPrefix(line, "?   \t") ||
-			strings.HasPrefix(line, "ok  ") || strings.HasPrefix(line, "?   ")
-		if packageSuccess && !strings.Contains(line, "coverage:") && !strings.Contains(line, "[no test files]") {
+		if omit(line) {
 			omitted++
 			continue
 		}
@@ -87,7 +282,7 @@ func reduceGoTestPassingPackages(output string) (string, int) {
 			break
 		}
 	}
-	summary := fmt.Sprintf("[zero] %d passing package lines omitted", omitted)
+	summary := fmt.Sprintf("[zero] %d %s omitted", omitted, label)
 	kept = append(kept, "")
 	copy(kept[insertAt+1:], kept[insertAt:])
 	kept[insertAt] = summary

internal/tools/shell_output_reducer_metrics_test.go

@@ -0,0 +1,107 @@
+package tools
+
+import (
+	"fmt"
+	"os"
+	"strings"
+	"testing"
+)
+
+type commandReducerCorpusCase struct {
+	name    string
+	command string
+	output  string
+	want    string
+}
+
+func commandReducerCorpus() []commandReducerCorpusCase {
+	goLines := []string{"output:"}
+	cargoTestLines := []string{"output:", "running 24 tests"}
+	cargoCheckLines := []string{"output:"}
+	pytestLines := []string{"output:"}
+	vitestLines := []string{"output:", " RUN  v3.2.4 /workspace"}
+	jestLines := []string{"output:"}
+	for index := 0; index < 24; index++ {
+		goLines = append(goLines, fmt.Sprintf("ok  \texample.test/pkg%d\t0.01s", index))
+		cargoTestLines = append(cargoTestLines, fmt.Sprintf("test module::case_%02d ... ok", index))
+		cargoCheckLines = append(cargoCheckLines, fmt.Sprintf("    Checking dependency_%02d v1.0.%d", index, index))
+		pytestLines = append(pytestLines, fmt.Sprintf("tests/test_module_%02d.py ........                         [%3d%%]", index, (index+1)*100/24))
+		vitestLines = append(vitestLines, fmt.Sprintf(" ✓ src/module_%02d.test.ts (8 tests) 12ms", index))
+		jestLines = append(jestLines, fmt.Sprintf("PASS src/module_%02d.test.ts", index))
+	}
+	goLines = append(goLines, "exit_code: 0")
+	cargoTestLines = append(cargoTestLines,
+		"test result: ok. 24 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.08s",
+		"exit_code: 0",
+	)
+	cargoCheckLines = append(cargoCheckLines,
+		"    Finished `dev` profile [unoptimized + debuginfo] target(s) in 2.41s",
+		"exit_code: 0",
+	)
+	pytestLines = append(pytestLines,
+		"============================= 192 passed in 3.21s =============================",
+		"exit_code: 0",
+	)
+	vitestLines = append(vitestLines,
+		" Test Files  24 passed (24)",
+		"      Tests  192 passed (192)",
+		"   Duration  3.21s",
+		"exit_code: 0",
+	)
+	jestLines = append(jestLines,
+		"Test Suites: 24 passed, 24 total",
+		"Tests:       192 passed, 192 total",
+		"Time:        3.21 s",
+		"exit_code: 0",
+	)
+	return []commandReducerCorpusCase{
+		{name: "go_test", command: "go test ./...", output: strings.Join(goLines, "\n"), want: "exit_code: 0"},
+		{name: "cargo_test", command: "cargo test --workspace", output: strings.Join(cargoTestLines, "\n"), want: "24 passed"},
+		{name: "cargo_check", command: "cargo check --workspace", output: strings.Join(cargoCheckLines, "\n"), want: "Finished `dev` profile"},
+		{name: "pytest", command: "pytest -q", output: strings.Join(pytestLines, "\n"), want: "192 passed"},
+		{name: "npm_vitest", command: "npm test", output: strings.Join(vitestLines, "\n"), want: "192 passed"},
+		{name: "pnpm_vitest", command: "pnpm test", output: strings.Join(vitestLines, "\n"), want: "192 passed"},
+		{name: "yarn_jest", command: "yarn test", output: strings.Join(jestLines, "\n"), want: "192 passed"},
+	}
+}
+
+func TestCommandOutputReductionCorpusMetrics(t *testing.T) {
+	t.Setenv("TMPDIR", t.TempDir())
+	totalRawTokens := 0
+	totalModelTokens := 0
+	for _, testCase := range commandReducerCorpus() {
+		t.Run(testCase.name, func(t *testing.T) {
+			result := reduceCommandOutput(ExecCommandToolName, map[string]any{"cmd": testCase.command}, Result{
+				Status: StatusOK,
+				Output: testCase.output,
+			})
+			if !strings.Contains(result.Output, testCase.want) {
+				t.Fatalf("decisive summary %q was lost: %q", testCase.want, result.Output)
+			}
+			if spillPath := result.Meta["spill_path"]; spillPath != "" {
+				spilled, err := os.ReadFile(spillPath)
+				if err != nil {
+					t.Fatalf("read raw artifact: %v", err)
+				}
+				if string(spilled) != testCase.output {
+					t.Fatal("raw artifact differs from original output")
+				}
+			}
+			rawTokens := estimateOutputTokens(testCase.output)
+			modelTokens := estimateOutputTokens(result.Output)
+			totalRawTokens += rawTokens
+			totalModelTokens += modelTokens
+			reduction := 100 * (rawTokens - modelTokens) / rawTokens
+			t.Logf("raw_bytes=%d model_bytes=%d raw_tokens=%d model_tokens=%d reduction_pct=%d reduced=%t",
+				len(testCase.output), len(result.Output), rawTokens, modelTokens, reduction, result.Meta["command_output_reduced"] == "true")
+		})
+	}
+	if totalRawTokens == 0 {
+		t.Fatal("empty reducer corpus")
+	}
+	if totalModelTokens >= totalRawTokens {
+		t.Fatalf("aggregate reducer output did not save tokens: raw=%d model=%d", totalRawTokens, totalModelTokens)
+	}
+	t.Logf("aggregate raw_tokens=%d model_tokens=%d reduction_pct=%d",
+		totalRawTokens, totalModelTokens, 100*(totalRawTokens-totalModelTokens)/totalRawTokens)
+}

internal/tools/shell_output_reducer_test.go

@@ -1,12 +1,32 @@
 package tools
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"strings"
 	"testing"
 )
 
+type commandReducerFixtureTool struct {
+	name   string
+	output string
+}
+
+func (tool commandReducerFixtureTool) Name() string { return tool.name }
+func (tool commandReducerFixtureTool) Description() string {
+	return "returns command output for reducer integration tests"
+}
+func (tool commandReducerFixtureTool) Parameters() Schema {
+	return Schema{Type: "object", AdditionalProperties: true}
+}
+func (tool commandReducerFixtureTool) Safety() Safety {
+	return Safety{SideEffect: SideEffectRead, Permission: PermissionAllow, Reason: "test output"}
+}
+func (tool commandReducerFixtureTool) Run(context.Context, map[string]any) Result {
+	return Result{Status: StatusOK, Output: tool.output}
+}
+
 func TestReduceCommandOutputCompactsPassingGoPackagesAndKeepsRawArtifact(t *testing.T) {
 	var lines []string
 	lines = append(lines, "output:")
@@ -131,3 +151,167 @@ func TestSelfManagedBudgetPreservesRawSpillWhenReducedOutputAlsoTruncates(t *tes
 		t.Fatalf("raw spill was not preserved: err=%v", err)
 	}
 }
+
+func TestReduceCommandOutputCompactsRecognizedTestAndBuildNoise(t *testing.T) {
+	t.Setenv("TMPDIR", t.TempDir())
+	for _, testCase := range commandReducerCorpus() {
+		t.Run(testCase.name, func(t *testing.T) {
+			result := reduceCommandOutput(ExecCommandToolName, map[string]any{"cmd": testCase.command}, Result{
+				Status: StatusOK,
+				Output: testCase.output,
+			})
+			if result.Meta["command_output_reduced"] != "true" {
+				t.Fatalf("recognized command output was not reduced: %#v", result)
+			}
+			if !strings.Contains(result.Output, testCase.want) || !strings.Contains(result.Output, "exit_code: 0") {
+				t.Fatalf("decisive summary or exit status was lost: %q", result.Output)
+			}
+			spillPath := result.Meta["spill_path"]
+			spilled, err := os.ReadFile(spillPath)
+			if err != nil {
+				t.Fatalf("read raw artifact: %v", err)
+			}
+			if string(spilled) != testCase.output {
+				t.Fatal("raw artifact differs from original output")
+			}
+		})
+	}
+}
+
+func TestReduceCommandOutputPreservesFailureEvidence(t *testing.T) {
+	t.Setenv("TMPDIR", t.TempDir())
+	passing := make([]string, 0, commandReducerMinPassingLines+2)
+	for index := 0; index < commandReducerMinPassingLines+2; index++ {
+		passing = append(passing, fmt.Sprintf("test module::case_%02d ... ok", index))
+	}
+	original := strings.Join(append(passing,
+		"test module::broken ... FAILED",
+		"failures:",
+		"---- module::broken stdout ----",
+		"assertion failed: expected zero",
+		"test result: FAILED. 14 passed; 1 failed; finished in 0.08s",
+		"exit_code: 101",
+	), "\n")
+
+	result := reduceCommandOutput(ExecCommandToolName, map[string]any{"cmd": "cargo test --workspace"}, Result{
+		Status: StatusError,
+		Output: original,
+	})
+	for _, evidence := range []string{"module::broken ... FAILED", "assertion failed: expected zero", "14 passed; 1 failed", "exit_code: 101"} {
+		if !strings.Contains(result.Output, evidence) {
+			t.Fatalf("failure evidence %q was lost: %q", evidence, result.Output)
+		}
+	}
+	if result.Meta["command_output_reduced"] != "true" {
+		t.Fatalf("repetitive passing output was not reduced: %#v", result)
+	}
+	spillPath := result.Meta["spill_path"]
+	if spillPath == "" {
+		t.Fatal("reduced failure output did not retain a raw artifact")
+	}
+	spilled, err := os.ReadFile(spillPath)
+	if err != nil {
+		t.Fatalf("read raw artifact: %v", err)
+	}
+	if string(spilled) != original {
+		t.Fatal("raw artifact differs from original output")
+	}
+}
+
+func TestReduceCommandOutputPassesThroughUnsupportedAndCompoundCommands(t *testing.T) {
+	outputs := make(map[string]string)
+	for _, fixture := range commandReducerCorpus() {
+		outputs[fixture.name] = fixture.output
+	}
+	testCases := []struct {
+		command string
+		output  string
+	}{
+		{command: "npm test | tee test.log", output: outputs["npm_vitest"]},
+		{command: "cargo test && echo done", output: outputs["cargo_test"]},
+		{command: "pytest; notify-send done", output: outputs["pytest"]},
+		{command: "cargo test 1>test.log", output: outputs["cargo_test"]},
+		{command: "cargo test 2>>errors.log", output: outputs["cargo_test"]},
+		{command: "git status --short", output: outputs["npm_vitest"]},
+	}
+	for _, testCase := range testCases {
+		result := reduceCommandOutput(ExecCommandToolName, map[string]any{"cmd": testCase.command}, Result{Status: StatusOK, Output: testCase.output})
+		if result.Output != testCase.output || result.Meta != nil {
+			t.Fatalf("command %q should pass through exactly, got %#v", testCase.command, result)
+		}
+	}
+}
+
+func TestReduceCommandOutputAllowsTrailingStderrMerge(t *testing.T) {
+	t.Setenv("TMPDIR", t.TempDir())
+	fixture := commandReducerCorpus()[1]
+	result := reduceCommandOutput(ExecCommandToolName, map[string]any{"cmd": fixture.command + " 2>&1"}, Result{
+		Status: StatusOK,
+		Output: fixture.output,
+	})
+	if result.Meta["command_output_reducer"] != "cargo_test" {
+		t.Fatalf("trailing stderr merge prevented reduction: %#v", result)
+	}
+}
+
+func TestReduceCommandOutputRequiresAuthoritativeRunnerSummary(t *testing.T) {
+	lookalikes := map[string]string{
+		"cargo test":  "test user::message ... ok",
+		"cargo check": "Checking a custom deployment state",
+		"pytest":      "application log ........ [100%]",
+		"npm test":    "PASS user-provided log message",
+	}
+	for command, line := range lookalikes {
+		original := strings.Repeat(line+"\n", commandReducerMinPassingLines+2) + "exit_code: 0"
+		result := reduceCommandOutput(ExecCommandToolName, map[string]any{"cmd": command}, Result{Status: StatusOK, Output: original})
+		if result.Output != original || result.Meta != nil {
+			t.Fatalf("summary-free output for %q should pass through, got %#v", command, result)
+		}
+	}
+}
+
+func TestReduceCommandOutputRejectsLookalikeRunnerSummaries(t *testing.T) {
+	t.Setenv("TMPDIR", t.TempDir())
+	testCases := []struct {
+		command string
+		line    string
+		summary string
+	}{
+		{command: "cargo check", line: "Compiling dependency v1.0.0", summary: "Finished generating bindings"},
+		{command: "pytest", line: "tests/test_module.py ........ [100%]", summary: "application log: 14 passed validation in staging"},
+		{command: "npm test", line: "PASS src/example.test.ts", summary: "Test Files discovered: 14"},
+	}
+	for _, testCase := range testCases {
+		original := strings.Repeat(testCase.line+"\n", commandReducerMinPassingLines+2) + testCase.summary + "\nexit_code: 1"
+		result := reduceCommandOutput(ExecCommandToolName, map[string]any{"cmd": testCase.command}, Result{Status: StatusError, Output: original})
+		if result.Output != original || result.Meta != nil {
+			t.Fatalf("lookalike summary for %q should pass through exactly, got %#v", testCase.command, result)
+		}
+	}
+}
+
+func TestRegistryAppliesCommandReducerToExecAndBashTools(t *testing.T) {
+	t.Setenv("TMPDIR", t.TempDir())
+	fixture := commandReducerCorpus()[1]
+	for _, toolName := range []string{ExecCommandToolName, "bash"} {
+		t.Run(toolName, func(t *testing.T) {
+			registry := NewRegistry()
+			registry.Register(commandReducerFixtureTool{name: toolName, output: fixture.output})
+			args := map[string]any{"cmd": fixture.command}
+			if toolName == "bash" {
+				args = map[string]any{"command": fixture.command}
+			}
+			result := registry.Run(context.Background(), toolName, args)
+			if result.Status != StatusOK || result.Meta["command_output_reducer"] != "cargo_test" {
+				t.Fatalf("registry result was not reduced: %#v", result)
+			}
+			if !strings.Contains(result.Output, fixture.want) {
+				t.Fatalf("registry reduction lost summary: %q", result.Output)
+			}
+			spilled, err := os.ReadFile(result.Meta["spill_path"])
+			if err != nil || string(spilled) != fixture.output {
+				t.Fatalf("registry raw artifact mismatch: err=%v", err)
+			}
+		})
+	}
+}

internal/tools/types.go

@@ -5,6 +5,7 @@ import (
 
 	"github.com/Gitlawb/zero/internal/execution"
 	"github.com/Gitlawb/zero/internal/sandbox"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
 )
 
 type SideEffect string
@@ -102,6 +103,16 @@ type Result struct {
 	// while callers migrate away from parsing text and string metadata.
 	ExecutionRequest *execution.Request `json:"-"`
 	ExecutionOutcome *execution.Outcome `json:"-"`
+	// Images are pictures the tool is handing the model — a screenshot it just
+	// captured, a file it was asked to look at. Text-only tools leave it nil.
+	//
+	// They are NOT delivered on this result's own tool-result message. Every
+	// provider drops images there: Anthropic maps a tool result to a tool_result
+	// block whose content is a string, Gemini to a functionResponse, and OpenAI
+	// guards its image content-parts to the user role. The agent loop therefore
+	// emits them as a following user message, which is also the only shape that
+	// keeps one tool result per tool call.
+	Images []zeroruntime.ImageBlock `json:"-"`
 	// Redacted is set when secret scrubbing altered Output before it left the
 	// tool-execution boundary.
 	Redacted bool

internal/tools/update_plan.go

@@ -26,7 +26,7 @@ func NewUpdatePlanTool() *updatePlanTool {
 	return &updatePlanTool{
 		baseTool: baseTool{
 			name: "update_plan",
-			description: "Create or update the in-memory plan for the current task. " +
+			description: "Create or update the in-memory plan for implementation or investigation with at least three meaningful dependent steps. Do not use it for simple lookups, explanations, or code navigation. " +
 				"Pass the full ordered list of steps each call; it replaces the previous plan. " +
 				"Each item needs a `content` string; `status` defaults to \"pending\" and `id` is " +
 				"auto-numbered, so you only need to supply `content` (and `status` as the task progresses). " +

internal/tools/view_image.go

@@ -0,0 +1,80 @@
+package tools
+
+import (
+	"context"
+
+	"github.com/Gitlawb/zero/internal/imageinput"
+	"github.com/Gitlawb/zero/internal/zeroruntime"
+)
+
+// ViewImageToolName is the canonical registry name of the image viewer.
+const ViewImageToolName = "view_image"
+
+// NewViewImageTool builds the workspace-scoped image viewer.
+func NewViewImageTool(workspaceRoot string) Tool {
+	return NewScopedViewImageTool(workspaceRoot, nil)
+}
+
+// NewScopedViewImageTool builds the image viewer bound to a path scope.
+//
+// The scope is not optional decoration. imageinput.LoadFile resolves a relative
+// path by joining it to the workspace root and otherwise takes the path as
+// given, so on its own it would happily read outside the workspace. Every path
+// goes through resolveScopedReadPath first, exactly like read_file, so this tool
+// grants no reach that read_file does not already have.
+func NewScopedViewImageTool(workspaceRoot string, scope PathScope) Tool {
+	return viewImageTool{
+		baseTool: baseTool{
+			name: ViewImageToolName,
+			description: "Look at an image file. Use this to see a screenshot, diagram, mockup or photo " +
+				"the user mentioned but did not attach, or one a previous tool wrote to disk. " +
+				"Accepts PNG, JPEG, GIF and WebP up to 10 MiB.",
+			parameters: Schema{
+				Type: "object",
+				Properties: map[string]PropertySchema{
+					"path": {Type: "string", Description: "Path of the image to look at."},
+				},
+				Required:             []string{"path"},
+				AdditionalProperties: false,
+			},
+			safety: readOnlySafety("Reads an image file without modifying anything."),
+			// Same shape as read_file: reads one file, mutates nothing, and is
+			// safe to run concurrently.
+			capabilities: ToolCapabilities{Effect: EffectReadOnly, ThreadSafe: true, ResourceKeys: fileResourceKeys},
+		},
+		workspaceRoot: workspaceRoot,
+		scope:         scope,
+	}
+}
+
+type viewImageTool struct {
+	baseTool
+	workspaceRoot string
+	scope         PathScope
+}
+
+func (tool viewImageTool) Run(_ context.Context, args map[string]any) Result {
+	requestedPath, err := aliasedStringArg(args, []string{"path", "file", "file_path", "filepath", "filename", "image"}, "", true, false)
+	if err != nil {
+		return errorResult("Error: Invalid arguments for " + ViewImageToolName + ": " + err.Error())
+	}
+	absolutePath, relativePath, err := resolveScopedReadPath(tool.workspaceRoot, tool.scope, requestedPath)
+	if err != nil {
+		return errorResult("Error: " + err.Error())
+	}
+	image, err := imageinput.LoadFile(absolutePath, tool.workspaceRoot)
+	if err != nil {
+		// LoadFile's errors already name the problem (missing, too large,
+		// unsupported type) and are safe to show: they carry the path the caller
+		// asked for, which it already knows.
+		return errorResult("Error viewing " + relativePath + ": " + err.Error())
+	}
+	return Result{
+		Status: StatusOK,
+		// The text says what arrived; the image itself reaches the model through
+		// the agent loop, which delivers Images on a following user message.
+		Output: "Viewing " + relativePath + " (" + image.MediaType + ").",
+		Images: []zeroruntime.ImageBlock{image},
+		Meta:   map[string]string{"media_type": image.MediaType},
+	}
+}

internal/tools/view_image_test.go

@@ -0,0 +1,100 @@
+package tools
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+// A 1x1 PNG. Small enough to inline, real enough to pass media-type sniffing.
+var onePixelPNG = []byte{
+	0x89, 'P', 'N', 'G', 0x0d, 0x0a, 0x1a, 0x0a,
+	0x00, 0x00, 0x00, 0x0d, 'I', 'H', 'D', 'R',
+	0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01,
+	0x08, 0x06, 0x00, 0x00, 0x00, 0x1f, 0x15, 0xc4,
+	0x89, 0x00, 0x00, 0x00, 0x0a, 'I', 'D', 'A', 'T',
+	0x78, 0x9c, 0x63, 0x00, 0x01, 0x00, 0x00, 0x05,
+	0x00, 0x01, 0x0d, 0x0a, 0x2d, 0xb4, 0x00, 0x00,
+	0x00, 0x00, 'I', 'E', 'N', 'D', 0xae, 0x42, 0x60, 0x82,
+}
+
+func TestViewImageReturnsTheImage(t *testing.T) {
+	root := t.TempDir()
+	if err := os.WriteFile(filepath.Join(root, "shot.png"), onePixelPNG, 0o600); err != nil {
+		t.Fatal(err)
+	}
+	result := NewViewImageTool(root).Run(context.Background(), map[string]any{"path": "shot.png"})
+
+	if result.Status != StatusOK {
+		t.Fatalf("status = %v: %s", result.Status, result.Output)
+	}
+	if len(result.Images) != 1 {
+		t.Fatalf("got %d images, want 1 — the text alone is useless to the model", len(result.Images))
+	}
+	if result.Images[0].MediaType != "image/png" {
+		t.Errorf("media type = %q, want image/png", result.Images[0].MediaType)
+	}
+	if len(result.Images[0].Data) != len(onePixelPNG) {
+		t.Errorf("image data is %d bytes, want %d", len(result.Images[0].Data), len(onePixelPNG))
+	}
+}
+
+// The tool must not reach outside the workspace. imageinput.LoadFile does no
+// containment check of its own, so the scoping is entirely this tool's job.
+func TestViewImageRefusesPathsOutsideTheWorkspace(t *testing.T) {
+	root := t.TempDir()
+	outside := t.TempDir()
+	secret := filepath.Join(outside, "outside.png")
+	if err := os.WriteFile(secret, onePixelPNG, 0o600); err != nil {
+		t.Fatal(err)
+	}
+	tool := NewViewImageTool(root)
+
+	for name, path := range map[string]string{
+		"absolute path outside": secret,
+		"parent traversal":      filepath.Join("..", filepath.Base(outside), "outside.png"),
+	} {
+		t.Run(name, func(t *testing.T) {
+			result := tool.Run(context.Background(), map[string]any{"path": path})
+			if result.Status == StatusOK {
+				t.Fatalf("read an image outside the workspace: %s", result.Output)
+			}
+			if len(result.Images) != 0 {
+				t.Error("a refused read still returned image bytes")
+			}
+		})
+	}
+}
+
+func TestViewImageRejectsNonImages(t *testing.T) {
+	root := t.TempDir()
+	if err := os.WriteFile(filepath.Join(root, "notes.txt"), []byte("not an image"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	result := NewViewImageTool(root).Run(context.Background(), map[string]any{"path": "notes.txt"})
+	if result.Status == StatusOK {
+		t.Fatalf("accepted a text file as an image: %s", result.Output)
+	}
+	if len(result.Images) != 0 {
+		t.Error("a rejected file still returned image bytes")
+	}
+}
+
+func TestViewImageReportsAMissingFileClearly(t *testing.T) {
+	result := NewViewImageTool(t.TempDir()).Run(context.Background(), map[string]any{"path": "nope.png"})
+	if result.Status == StatusOK {
+		t.Fatal("a missing file reported success")
+	}
+	// The error names the path the caller asked for, matching read_file. Both
+	// also carry the resolved absolute path from the underlying syscall error;
+	// that is pre-existing behaviour shared with read_file, not something this
+	// tool introduces, so it is not asserted either way here.
+	if !strings.Contains(result.Output, "nope.png") {
+		t.Errorf("error should name the file, got %q", result.Output)
+	}
+	if len(result.Images) != 0 {
+		t.Error("a failed read still returned image bytes")
+	}
+}

internal/tui/clipboard_nul_test.go

@@ -0,0 +1,60 @@
+package tui
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/charmbracelet/x/ansi"
+)
+
+// A COPIED NUL KILLED THE WHOLE PROGRAM ON WINDOWS.
+//
+// github.com/atotto/clipboard's Windows path calls syscall.StringToUTF16, which
+// PANICS rather than returning an error when the string contains a NUL — so
+// selecting transcript text that carried one took the TUI down with "program
+// experienced a panic" and a stack ending in clipboard.writeAll. A teammate hit
+// it on a real Windows machine.
+//
+// The NUL reaches the clipboard because ansi.Strip removes escape SEQUENCES and
+// leaves C0 bytes alone, and the transcript carries them from two directions:
+// tool output (binary reads, git's -z listings) and this package's own card
+// protocol, whose prefixes are literally "\x00command-card\x00".
+func TestNulNeverReachesTheClipboard(t *testing.T) {
+	// The card prefix is a real, in-repo source: ansi.Strip leaves it intact.
+	if !strings.ContainsRune(ansi.Strip(commandCardTranscriptPrefix+"Profile"), 0) {
+		t.Fatal("setup: the card prefix no longer carries a NUL; this test guards the wrong thing")
+	}
+	for _, tc := range []struct{ name, in string }{
+		{"card prefix", commandCardTranscriptPrefix + "Profile\nstatus: ok"},
+		{"plan card prefix", planCardTranscriptPrefix + "plan"},
+		{"binary tool output", "before\x00after"},
+		{"git -z listing", "a.go\x00b.go\x00c.go"},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			got := clipboardSafeText(tc.in)
+			if strings.ContainsRune(got, 0) {
+				t.Fatalf("a NUL survived to the clipboard write: %q — Windows would panic here", got)
+			}
+			// The readable content must survive: this sanitises, it does not truncate.
+			for _, want := range strings.Split(strings.ReplaceAll(tc.in, "\x00", "\n"), "\n") {
+				if want = strings.TrimSpace(want); want != "" && !strings.Contains(got, want) {
+					t.Fatalf("sanitising dropped real content %q from %q", want, got)
+				}
+			}
+		})
+	}
+}
+
+// Newlines and tabs are legal clipboard content and must survive — a multi-line
+// selection is the normal case, not the exception.
+func TestClipboardSanitisingKeepsLayout(t *testing.T) {
+	in := "line one\n\tindented\nline three"
+	if got := clipboardSafeText(in); got != in {
+		t.Fatalf("clean multi-line text was altered:\n got %q\nwant %q", got, in)
+	}
+	// Other C0 bytes go: they would terminate the OSC52 escape early and
+	// corrupt the copy even where they do not panic.
+	if got := clipboardSafeText("a\x07b\x1bc"); got != "abc" {
+		t.Fatalf("control bytes survived: %q", got)
+	}
+}

internal/tui/transcript_selection.go

@@ -1499,7 +1499,54 @@ func (m model) selectedTranscriptText() string {
 	return strings.Join(parts, "\n")
 }
 
+// clipboardSafeText removes the control characters a clipboard write cannot
+// carry, keeping newlines and tabs.
+//
+// A NUL PANICS THE WHOLE PROGRAM ON WINDOWS, and that is not a theoretical
+// concern: syscall.StringToUTF16 — which github.com/atotto/clipboard calls on
+// Windows — PANICS rather than erroring when the string contains one, so a
+// single copied NUL killed the TUI with "program experienced a panic". It
+// reached the clipboard because ansi.Strip removes escape SEQUENCES and leaves
+// C0 bytes untouched, and the transcript legitimately carries them from two
+// directions: tool output (a read of a binary file, git's -z NUL-separated
+// listings) and this package's own card protocol, whose row prefixes are
+// literally "\x00command-card\x00".
+//
+// Stripped at the WRITE, not at each source, because the sources are unbounded:
+// any tool result a user can select from is a source, and a fix per source is
+// one that the next tool re-breaks. The same pass protects the OSC52 fallback,
+// where a stray control byte would terminate the escape sequence early and
+// corrupt the copy rather than crash.
+//
+// \n and \t survive: a multi-line selection is the normal case, and both are
+// legal clipboard content.
+func clipboardSafeText(text string) string {
+	if !strings.ContainsFunc(text, unsafeClipboardRune) {
+		return text
+	}
+	var out strings.Builder
+	out.Grow(len(text))
+	for _, r := range text {
+		if unsafeClipboardRune(r) {
+			continue
+		}
+		out.WriteRune(r)
+	}
+	return out.String()
+}
+
+func unsafeClipboardRune(r rune) bool {
+	if r == '\n' || r == '\t' {
+		return false
+	}
+	return r < 0x20 || r == 0x7f
+}
+
 func copyTranscriptSelectionCmd(text string) tea.Cmd {
+	// SANITISED ONCE, HERE, before either destination sees it — the native
+	// clipboard would panic on a NUL and the OSC52 fallback would be corrupted
+	// by one.
+	text = clipboardSafeText(text)
 	return func() tea.Msg {
 		// Prefer the native OS clipboard (pbcopy / clip.exe / xclip): it works on
 		// local terminals — including macOS Terminal.app, which has no OSC52 support