@harness-engineering/cli 1.14.0 → 1.16.0

package/dist/agents/commands/cursor/harness/harness-verification.mdc

@@ -0,0 +1,429 @@
+---
+description: Comprehensive harness verification of project health and compliance
+alwaysApply: false
+---
+
+<!-- Generated by harness generate-slash-commands. Do not edit. -->
+
+# Harness Verification
+
+> 3-level evidence-based verification. No completion claims without fresh evidence. "Should work" is not evidence.
+
+## When to Use
+
+- After completing any implementation task (before claiming "done")
+- After executing a plan or spec (verify all deliverables)
+- When validating work done by another agent or in a previous session
+- When resuming work after a context reset (re-verify before continuing)
+- When `on_commit` or `on_pr` triggers fire and verification is needed
+- NOT as a replacement for tests (verification checks that tests exist and pass, not that logic is correct)
+- NOT for in-progress work (verify at completion boundaries, not mid-stream)
+
+### Verification Tiers
+
+Harness uses a two-tier verification model:
+
+| Tier           | Skill                             | When                       | What                                               |
+| -------------- | --------------------------------- | -------------------------- | -------------------------------------------------- |
+| **Quick gate** | harness-execution (built-in)      | After every task           | test + lint + typecheck + build + harness validate |
+| **Deep audit** | harness-verification (this skill) | Milestones, PRs, on-demand | EXISTS → SUBSTANTIVE → WIRED                       |
+
+Use this skill (deep audit) for milestone boundaries, before creating PRs, or when the quick gate passes but something feels wrong. Do NOT invoke this skill after every individual task — that is what the quick gate handles.
+
+## Process
+
+### Iron Law
+
+**No completion claim may be made without fresh verification evidence collected in THIS session.**
+
+Cached results, remembered outcomes, and "it worked last time" are not evidence. Run the checks. Read the output. Report what you observed.
+
+The words "should", "probably", "seems to", and "I believe" are forbidden in verification reports. Replace with "verified: [evidence]" or "not verified: [what is missing]."
+
+---
+
+### Context Loading
+
+Before running verification levels, load session context if a session slug was provided (e.g., by autopilot dispatch):
+
+```json
+gather_context({
+  path: "<project-root>",
+  intent: "Verify phase deliverables",
+  skill: "harness-verification",
+  session: "<session-slug-if-provided>",
+  include: ["state", "learnings", "validation"]
+})
+```
+
+**Session resolution:** If a session slug is known (passed via autopilot dispatch or available from a previous handoff), include the `session` parameter. This scopes all state reads to `.harness/sessions/<slug>/`. If no session is known, omit it — `gather_context` falls back to global files at `.harness/`.
+
+Use the returned learnings to check for known failures and dead ends relevant to the artifacts being verified.
+
+---
+
+### Level 1: EXISTS — The Artifact Is Present
+
+For every artifact that was supposed to be created or modified:
+
+1. **Check that the file exists on disk.** Use `ls`, `stat`, or read the file. Do not assume it exists because you wrote it — file writes can fail silently.
+
+2. **Check that the file has content.** An empty file is not an artifact. Read the file and confirm it has non-trivial content.
+
+3. **Check the file is in the right location.** Compare the actual path against the spec or plan. A file in the wrong directory is not "present."
+
+4. **Record the result.** For each expected artifact:
+   ```
+   [EXISTS: PASS] path/to/file.ts (247 lines)
+   [EXISTS: FAIL] path/to/missing-file.ts — file not found
+   ```
+
+Do not proceed to Level 2 until all Level 1 checks pass. Missing files must be created first.
+
+---
+
+### Level 2: SUBSTANTIVE — Not a Stub
+
+For every artifact that passed Level 1:
+
+1. **Read the file content.** Do not skim — read it thoroughly.
+
+2. **Scan for anti-patterns** that indicate stub or placeholder implementations:
+   - `TODO` or `FIXME` comments (especially `TODO: implement`)
+   - `throw new Error('not implemented')`
+   - `() => {}` (empty arrow functions)
+   - `return null`, `return undefined`, `return {}` as the only logic
+   - `pass` (Python placeholder)
+   - `placeholder`, `stub`, `mock` in non-test code
+   - Functions with only a comment describing what they should do
+   - Interfaces or types defined but never implemented
+
+3. **Verify real implementation exists.** The file must contain actual logic that performs the described behavior. A function that only returns a hardcoded value is a stub unless that is the correct behavior.
+
+4. **Check for completeness against the spec.** If the spec says "handles errors X, Y, Z," verify all three are handled, not just X.
+
+5. **Record the result.** For each artifact:
+   ```
+   [SUBSTANTIVE: PASS] path/to/file.ts — real implementation, no stubs
+   [SUBSTANTIVE: FAIL] path/to/file.ts — contains TODO on line 34, empty handler on line 67
+   ```
+
+Do not proceed to Level 3 until all Level 2 checks pass. Stubs must be replaced with real implementations first.
+
+---
+
+### Level 3: WIRED — Connected to the System
+
+For every artifact that passed Level 2:
+
+1. **Verify the artifact is imported/required** by at least one other file in the system (unless it is an entry point).
+
+2. **Verify the artifact is called/used.** An import that is never called is dead code. Trace the usage:
+   - Functions: called from at least one other function or test
+   - Components: rendered in at least one parent or route
+   - Types: used in at least one function signature or variable declaration
+   - Configuration: loaded and applied by the system
+   - Tests: executed by the test runner
+
+3. **Verify the artifact is tested.** There must be at least one test that exercises the artifact's behavior. Check:
+   - Test file exists
+   - Test imports or references the artifact
+   - Test makes assertions about the artifact's behavior
+   - Test actually runs (not skipped with `.skip` or `xit`)
+
+4. **Run the tests.** Execute the test suite and verify tests pass. Do not trust "they passed earlier" — run them now.
+
+5. **Run harness checks.** Execute `harness validate` and verify the artifact integrates correctly with the project's constraints.
+
+6. **Record the result.** For each artifact:
+   ```
+   [WIRED: PASS] path/to/file.ts — imported by 3 files, tested in file.test.ts (4 tests, all pass)
+   [WIRED: FAIL] path/to/file.ts — exported but not imported by any other file
+   ```
+
+---
+
+### Anti-Pattern Scan
+
+Run this scan across all changed files as a final check:
+
+```
+Scan targets: TODO, FIXME, XXX, HACK, PLACEHOLDER, NOT_IMPLEMENTED
+Code patterns: () => {}, return null (as sole body), pass, raise NotImplementedError
+Test patterns: .skip, xit, xdescribe, @pytest.mark.skip, pending
+```
+
+Any match is a verification failure. Either fix it or explicitly document why it is acceptable (e.g., "TODO is tracked in issue #123 and out of scope for this task").
+
+---
+
+### Gap Identification
+
+After running all three levels, produce a structured gap report:
+
+```
+## Verification Report
+
+### Level 1: EXISTS
+- [PASS] path/to/file-a.ts (120 lines)
+- [PASS] path/to/file-b.ts (85 lines)
+- [FAIL] path/to/file-c.ts — not found
+
+### Level 2: SUBSTANTIVE
+- [PASS] path/to/file-a.ts — real implementation
+- [FAIL] path/to/file-b.ts — TODO on line 22
+
+### Level 3: WIRED
+- [PASS] path/to/file-a.ts — imported, tested, harness passes
+- [NOT CHECKED] path/to/file-b.ts — blocked by Level 2 failure
+
+### Anti-Pattern Scan
+- path/to/file-b.ts:22 — TODO: implement validation
+
+### Gaps
+1. path/to/file-c.ts must be created
+2. path/to/file-b.ts:22 must be implemented (not stub)
+
+### Verdict: INCOMPLETE — 2 gaps must be resolved
+```
+
+The verification report uses conventional markdown patterns for structured output:
+
+```
+**[CRITICAL]** path/to/file.ts:22 — TODO: implement validation (anti-pattern)
+**[IMPORTANT]** path/to/file.ts — exported but not imported by any other file
+```
+
+### Verification Sign-Off
+
+After producing the verification report, request acceptance:
+
+```json
+emit_interaction({
+  path: "<project-root>",
+  type: "confirmation",
+  confirmation: {
+    text: "Verification report: <VERDICT>. Accept and proceed?",
+    context: "<summary: N artifacts checked, N gaps found>",
+    impact: "Accepting proceeds to code review. Declining requires gap resolution first.",
+    risk: "<low if PASS, high if gaps remain>"
+  }
+})
+```
+
+### Handoff and Transition
+
+After producing the verification report, write the handoff and conditionally transition:
+
+Write `.harness/handoff.json`:
+
+```json
+{
+  "fromSkill": "harness-verification",
+  "phase": "COMPLETE",
+  "summary": "<verdict summary>",
+  "artifacts": ["<verified file paths>"],
+  "verdict": "pass | fail",
+  "gaps": ["<gap descriptions if any>"]
+}
+```
+
+**Write session summary (if session is known).** If running within a session context, update the session summary:
+
+```json
+writeSessionSummary(projectPath, sessionSlug, {
+  session: "<session-slug>",
+  lastActive: "<ISO timestamp>",
+  skill: "harness-verification",
+  status: "Verification <PASS|FAIL|INCOMPLETE>. <N> artifacts checked, <N> gaps.",
+  spec: "<spec path if known>",
+  keyContext: "<1-2 sentences: verification outcome, any gaps found>",
+  nextStep: "<e.g., Proceed to code review / Resolve gaps>"
+})
+```
+
+If no session slug is known, skip this step.
+
+**If verdict is PASS (all levels passed, no gaps):**
+
+Call `emit_interaction`:
+
+```json
+{
+  "type": "transition",
+  "transition": {
+    "completedPhase": "verification",
+    "suggestedNext": "review",
+    "reason": "Verification passed at all 3 levels",
+    "artifacts": ["<verified file paths>"],
+    "requiresConfirmation": false,
+    "summary": "Verification passed: <N> artifacts checked. EXISTS, SUBSTANTIVE, WIRED all passed.",
+    "qualityGate": {
+      "checks": [
+        { "name": "level1-exists", "passed": true, "detail": "<N> artifacts present" },
+        { "name": "level2-substantive", "passed": true, "detail": "No stubs or placeholders" },
+        {
+          "name": "level3-wired",
+          "passed": true,
+          "detail": "All artifacts imported, tested, integrated"
+        },
+        { "name": "anti-pattern-scan", "passed": true, "detail": "No matches" },
+        { "name": "harness-validate", "passed": true }
+      ],
+      "allPassed": true
+    }
+  }
+}
+```
+
+The response will include `nextAction: "Invoke harness-code-review skill now"`.
+Immediately invoke harness-code-review without waiting for user input.
+
+**If verdict is FAIL or INCOMPLETE:**
+
+Do NOT emit a transition. Surface gaps to the user for resolution. The handoff is written with the gaps recorded for future reference.
+
+---
+
+### Regression Test Verification
+
+When verifying a bug fix, apply this extended protocol:
+
+1. **Write** the regression test that reproduces the bug
+2. **Run** the test — it must PASS (proving the fix works)
+3. **Revert** the fix (temporarily): `git stash` or comment out the fix
+4. **Run** the test — it must FAIL (proving the test actually catches the bug)
+5. **Restore** the fix: `git stash pop` or uncomment
+6. **Run** the test — it must PASS again (proving the fix is the reason)
+
+If step 4 passes (test does not fail without the fix), the test is not a valid regression test. It does not catch the bug. Rewrite it.
+
+## Evidence Requirements
+
+This skill is the primary evidence producer in the workflow. Every pass/fail assertion in the verification report MUST include concrete evidence. The words "should", "probably", and "seems to" are already forbidden by the Iron Law -- this section defines HOW to cite evidence.
+
+Every verification claim MUST use one of:
+
+1. **File reference:** `file:line` format with observed content (e.g., `src/services/user-service.ts:42` -- "create method validates email format before insert")
+2. **Test output:** Include the actual test command and its complete output:
+   ```
+   $ npx vitest run src/services/user-service.test.ts
+   PASS  src/services/user-service.test.ts
+     UserService
+       create (4 tests)
+       list (3 tests)
+       expiry (2 tests)
+   Tests: 9 passed, 9 total
+   ```
+3. **Harness output:** Include full `harness validate` and `harness check-deps` output
+4. **Anti-pattern scan output:** Include the actual grep/search command and results (or absence of results)
+5. **Import chain evidence:** Include the actual import statements found when verifying WIRED level
+6. **Session evidence:** Write to the `evidence` session section for each verification level:
+   ```json
+   manage_state({
+     action: "append_entry",
+     session: "<current-session>",
+     section: "evidence",
+     authorSkill: "harness-verification",
+     content: "[EXISTS:PASS] src/services/user-service.ts (189 lines) -- verified via direct file read"
+   })
+   ```
+
+**When to cite:** At every verification level. Level 1 (EXISTS) cites file reads. Level 2 (SUBSTANTIVE) cites specific line content. Level 3 (WIRED) cites import statements, test execution output, and harness check output. The verification report format already requires `[PASS]`/`[FAIL]` markers -- each marker must be accompanied by the evidence that produced it.
+
+**Uncited claims:** ANY verification assertion without direct evidence is a verification failure, not merely an uncited claim. This skill does not use `[UNVERIFIED]` -- if evidence cannot be produced, the verdict is FAIL or INCOMPLETE.
+
+## Harness Integration
+
+- **`harness validate`** — Run in Level 3 WIRED check. Verifies project-wide health and constraint compliance.
+- **`harness check-deps`** — Run in Level 3 to verify new artifacts respect dependency boundaries.
+- **`harness check-docs`** — Run to verify documentation is updated for new artifacts. Missing docs for new public APIs is a gap.
+- **Test runner** — Must be run fresh (not cached) during Level 3. Read actual output, check exit codes.
+
+All commands must be run fresh in the current session. Do not rely on results from a previous session or a previous run in the same session if code has changed since.
+
+- **`emit_interaction`** -- Call after verification passes to auto-transition to harness-code-review. Only emitted on PASS verdict. Uses auto-transition (proceeds immediately).
+
+## Success Criteria
+
+- Every claimed deliverable has been verified at all 3 levels
+- No anti-patterns remain in delivered code
+- Verification report uses the structured format with PASS/FAIL per artifact per level
+- All verification evidence was collected fresh in the current session
+- No forbidden language ("should", "probably", "seems to") appears in the report
+- All gaps are explicitly identified with specific remediation steps
+- Regression tests (for bug fixes) pass the 5-step revert check
+
+## Non-Determinism Tolerance
+
+For mechanical checks (tests pass, lint clean, types check), results are binary — pass or fail. No tolerance.
+
+For behavioral verification (did the agent follow a convention, did the output match a style guide), accept threshold-based results:
+
+- Run the check multiple times if needed
+- "Agent followed the constraint in 4/5 runs" = pass
+- "Agent followed the constraint in 2/5 runs" = fail — the convention is poorly written, not the agent
+
+If a behavioral convention fails more than 40% of the time, the convention needs rewriting. Blame the instruction, not the executor.
+
+## Examples
+
+### Example: Verifying a New Service Module
+
+Task: "Create UserService with create, read, update, delete operations."
+
+```
+## Verification Report
+
+### Level 1: EXISTS
+- [PASS] src/services/user-service.ts (189 lines)
+- [PASS] src/services/user-service.test.ts (245 lines)
+- [PASS] src/services/index.ts (updated — exports UserService)
+
+### Level 2: SUBSTANTIVE
+- [PASS] src/services/user-service.ts — all 4 CRUD methods implemented with
+         validation, error handling, and database calls
+- [PASS] src/services/user-service.test.ts — 12 tests covering happy paths,
+         error cases, and edge cases (no skipped tests)
+
+### Level 3: WIRED
+- [PASS] src/services/user-service.ts — imported by src/api/routes/users.ts,
+         tested in user-service.test.ts (12 tests, all pass)
+- [PASS] harness validate — passes
+- [PASS] harness check-deps — no boundary violations
+
+### Anti-Pattern Scan
+- No matches found
+
+### Gaps
+(none)
+
+### Verdict: COMPLETE — all artifacts verified at all levels
+```
+
+## Gates
+
+- **No completion without evidence.** You may not say "done," "complete," "finished," or "implemented" without a verification report showing PASS at all 3 levels for all deliverables.
+- **No stale evidence.** Evidence must be from the current session. "I checked earlier" is not evidence. Run it again.
+- **No forbidden language.** "Should work," "probably fine," "seems correct," and "I believe it works" are not verification statements. Replace with observed evidence or state "not verified."
+- **No skipping levels.** Level 1 before Level 2. Level 2 before Level 3. Each level depends on the previous.
+- **No satisfaction before evidence.** The natural inclination after writing code is to feel done. Resist it. Feeling done is not being done. Evidence is being done.
+
+## Escalation
+
+- **When an artifact cannot pass Level 3 (WIRED) because the system it connects to does not exist yet:** Document the gap explicitly. State what integration is missing and what must be built. Do not mark it as PASS.
+- **When anti-pattern scan finds TODOs that are intentional:** Each must be justified with a tracked issue number. "TODO: implement" with no issue reference is not acceptable. "TODO(#123): add rate limiting after infrastructure is ready" is acceptable.
+- **When tests pass but you suspect they are not testing real behavior:** Read the test assertions carefully. If tests only check "does not throw" or assert on mock return values without verifying real behavior, flag them as SUBSTANTIVE failures.
+- **When verification reveals the spec itself is incomplete:** Do not fill in the gaps yourself. Escalate to the human: "Verification found that the spec does not define behavior for [scenario]. How should this be handled?"
+- **When you cannot run harness checks:** If `harness validate` or `harness check-deps` cannot be run (missing configuration, broken tooling), this is a blocking issue. Do not skip verification — fix the tooling or escalate.
+
+## Harness Integration
+
+- **`gather_context`** — Used in Context Loading phase (before Level 1) to load session-scoped state, learnings, and validation in a single call. The `session` parameter scopes reads to the session directory when provided by autopilot dispatch.
+- **`harness validate`** — Run during Level 3 (WIRED) to verify artifact integration.
+- **`harness check-deps`** — Run during Level 3 (WIRED) to verify dependency boundaries.
+
+After verification completes, append a tagged learning:
+
+- **YYYY-MM-DD [skill:harness-verification] [outcome:pass/fail]:** Verified [feature]. [Brief note on what was found or confirmed.]
+

package/dist/agents/commands/cursor/harness/harness-verify.mdc

@@ -0,0 +1,167 @@
+---
+description: Binary pass/fail quick gate — runs test, lint, typecheck commands and returns structured result
+alwaysApply: false
+---
+
+<!-- Generated by harness generate-slash-commands. Do not edit. -->
+
+# Harness Verify
+
+> Binary pass/fail quick gate. Runs test, lint, typecheck — returns structured result. No judgment calls, no deep analysis. Pass or fail.
+
+## When to Use
+
+- After every task as a final sanity check
+- As the final step in any code-producing skill
+- When you need a fast mechanical answer: "does the code compile, lint, and pass tests?"
+- NOT for deep verification or architectural review (use `harness-verification` for that)
+- NOT for AI-powered code review (use `harness-code-review` for that)
+
+## Relationship to Other Skills
+
+| Skill                     | What It Does                                   | Time  |
+| ------------------------- | ---------------------------------------------- | ----- |
+| **harness-verify** (this) | Mechanical quick gate: typecheck, lint, test   | ~30s  |
+| **harness-verification**  | Deep audit: architecture, patterns, edge cases | ~5min |
+
+`harness-verify` is the fast, deterministic gate. `harness-verification` is the slow, thorough audit. They serve different purposes and should not be confused.
+
+## Process
+
+### Phase 1: DETECT
+
+Auto-detect project commands by inspecting the project root:
+
+1. **package.json** — Look for `scripts.test`, `scripts.lint`, `scripts.typecheck` (or `scripts.tsc`, `scripts.type-check`)
+2. **Makefile** — Look for `test`, `lint`, `typecheck` targets
+3. **Conventions** — Fall back to common commands:
+   - Typecheck: `npx tsc --noEmit`, `mypy .`, `go vet ./...`
+   - Lint: `npx eslint .`, `ruff check .`, `golangci-lint run`
+   - Test: `npm test`, `pytest`, `go test ./...`
+4. **Language detection** — Use file extensions to determine which convention set applies
+
+For each of the three checks (typecheck, lint, test), record either the detected command or `NONE` if no command can be determined.
+
+### Phase 2: EXECUTE
+
+Run all detected commands in this order: **typecheck -> lint -> test**.
+
+Rules:
+
+- Run ALL commands regardless of earlier failures. Do not short-circuit.
+- Capture exit code, stdout, and stderr for each command.
+- Exit code 0 = PASS. Any non-zero exit code = FAIL.
+- If a command was `NONE` (not detected), mark that check as `SKIPPED`.
+- Run each command from the project root directory.
+- Do not modify any files. Do not install dependencies. Do not fix errors.
+
+### Design Constraint Check (conditional)
+
+When `harness.config.json` contains a `design` block:
+
+1. **Run design constraint checks** by invoking `harness-accessibility` in scan+evaluate mode against the project.
+2. Apply the `design.strictness` setting to determine severity:
+   - `strict`: accessibility violations are FAIL; anti-pattern violations are WARN
+   - `standard`: accessibility and anti-pattern violations are WARN; nothing blocks
+   - `permissive`: all design violations are INFO
+3. Capture the result as `Design: [PASS/WARN/FAIL/SKIPPED]`.
+4. If no `design` block exists in config, mark Design as `SKIPPED`.
+
+The design check runs AFTER test/lint/typecheck. It does not short-circuit on earlier failures.
+
+### Phase 3: REPORT
+
+Output a structured result in this exact format:
+
+```
+Verification: [PASS/FAIL]
+- Typecheck: [PASS/FAIL/SKIPPED]
+- Lint:      [PASS/FAIL/SKIPPED]
+- Test:      [PASS/FAIL/SKIPPED]
+```
+
+When design config is present, include the design line:
+
+```
+Verification: [PASS/FAIL]
+- Typecheck: [PASS/FAIL/SKIPPED]
+- Lint:      [PASS/FAIL/SKIPPED]
+- Test:      [PASS/FAIL/SKIPPED]
+- Design:    [PASS/WARN/FAIL/SKIPPED]
+```
+
+Rules:
+
+- Overall `Verification: PASS` only if all non-skipped checks passed.
+- If all checks are SKIPPED, overall result is `PASS` (nothing to fail).
+- On FAIL, include a brief summary of what failed (e.g., "3 type errors", "2 lint errors", "5 tests failed") below the structured block.
+
+### Roadmap Sync (conditional)
+
+When all non-skipped checks pass (overall `Verification: PASS`) and `docs/roadmap.md` exists:
+
+1. Trigger a roadmap sync to update feature statuses based on the verified state.
+2. Use the `manage_roadmap` MCP tool with `sync` action if available, or note to the caller that a roadmap sync is recommended.
+3. Features linked to plans whose tasks are all complete and verified may be marked as `done`.
+
+If `docs/roadmap.md` does not exist, skip this step silently. If verification failed, do not sync — the roadmap should only reflect verified completions.
+
+## Deterministic Checks
+
+This skill is entirely deterministic. There are no LLM judgment calls anywhere in the process.
+
+- Exit code 0 = PASS. Always.
+- Exit code non-zero = FAIL. Always.
+- No "it looks like this might be okay" reasoning. No interpretation of output.
+- The same codebase with the same commands will always produce the same result.
+
+## Harness Integration
+
+- Follows Principle 7 (Deterministic-vs-LLM Split) — this skill is entirely deterministic
+- Invoked as the final step by code-producing skills (harness-execution, harness-tdd)
+- Complements harness-verification (deep audit) — use verify for quick checks, verification for milestones
+- Output format is consumed by harness-integrity for the unified pipeline
+- Invokes `harness-accessibility` for design constraint checking when `design` config exists
+- Design violations respect `design.strictness` from `harness.config.json`
+- **Roadmap sync** — When verification passes and `docs/roadmap.md` exists, triggers `manage_roadmap sync` to mark verified features as `done`. Only fires on overall PASS.
+
+## Success Criteria
+
+- [ ] All detected commands were executed
+- [ ] Report follows the structured format exactly
+- [ ] Overall verdict correctly reflects individual results
+- [ ] Failed checks include error output summary
+
+## Examples
+
+### Example: Node.js Project
+
+```
+Verification: PASS
+- Types: PASS (no errors)
+- Lint: PASS (0 warnings)
+- Tests: PASS (42/42)
+```
+
+### Example: Failing Project
+
+```
+Verification: FAIL
+- Types: FAIL (3 type errors in src/auth/login.ts)
+- Lint: PASS
+- Tests: NOT RUN
+```
+
+## Gates
+
+- **No judgment calls.** The exit code is the only signal.
+- **No skipping.** If a command is detected, it runs. Period.
+- **Fresh execution only.** Do not cache results. Do not reuse previous runs. Execute the commands right now.
+- **No file modifications.** This skill is read-only (plus command execution). It must not change the codebase.
+
+## Escalation
+
+- **Timeout:** Each command has a 120-second timeout. If a command exceeds this, mark it as FAIL with reason "TIMEOUT".
+- **No commands detected:** If no typecheck, lint, or test commands can be detected, all three checks are SKIPPED and the overall result is PASS. Log a note that no verification commands were found.
+- **Environment errors:** If a command fails due to missing tooling (e.g., `tsc` not installed), mark it as FAIL. Do not attempt to install the tooling.
+