jh941213 Code Verification
Identical to the verification the TaskCompleted hook (
verify-task-quality.sh) runs automatically. Use this skill when the hook does not fire (manual verification requests, environments without the hook installed).
Verifies code after task completion. Run in order; proceed to the next step only after the current one passes.
Step 1: Project Detection
| File | Stack | Typecheck | Lint | Test | Build |
|---|---|---|---|---|---|
package.json |
Node/TS | npx tsc --noEmit |
npx eslint . / npx biome check . |
npx vitest run / npx jest |
npm run build |
pyproject.toml |
Python | mypy . / pyright . |
ruff check . |
pytest -q |
— |
go.mod |
Go | go vet ./... |
golangci-lint run |
go test ./... |
go build ./... |
Cargo.toml |
Rust | — | cargo clippy |
cargo test |
cargo build |
# Auto-detect
[ -f package.json ] && cat package.json | python3 -c "import sys,json; s=json.load(sys.stdin).get('scripts',{}); print('\n'.join(f'{k}: {v}' for k,v in s.items()))"
[ -f pyproject.toml ] && echo "Python project detected"
[ -f go.mod ] && echo "Go project detected"
[ -f Cargo.toml ] && echo "Rust project detected"
Step 2: Verification Order (fix immediately on failure)
2-1. TypeScript Typecheck
npx tsc --noEmit
On failure: fix type errors → re-run. After 3 failures, escalate to the user.
2-2. Lint
# ESLint
npx eslint . --max-warnings=0
# or Biome
npx biome check .
# Python
ruff check .
On failure: try --fix auto-fix → manual fix → re-run.
2-3. Tests
# Vitest
npx vitest run
# Jest
npx jest --passWithNoTests
# pytest
pytest -q
On failure: analyze failing tests → fix the code (modifying tests is a last resort) → re-run.
2-4. Build
npm run build
On failure: analyze build errors → fix → re-run.
2-5. Circular Dependency Detection (JS/TS projects)
npx madge --circular --extensions ts,tsx src/ 2>/dev/null
If circular dependencies are found: report as WARNING. Only newly introduced cycles are fix targets.
2-6. Dead Code Detection (JS/TS projects)
npx knip --no-exit-code 2>/dev/null | head -50
If unused exports/files/dependencies are found: report as INFO. Do not auto-fix.
2-7. Secret Scan
gitleaks detect --source . --no-git -v 2>&1 | head -20
If secrets are found: CRITICAL. Report immediately and drive a fix.
2-8. AI Slop Pattern Detection (ast-grep)
# Unnecessary console.log
sg --pattern 'console.log($$$)' --lang ts src/ 2>/dev/null | head -10
# any type usage
sg --pattern '$A as any' --lang ts src/ 2>/dev/null | head -10
If found: report as WARNING.
Step 3: Result Report
| Step | Status | Details |
|---|---|---|
| typecheck | PASS/FAIL/SKIP | Error count or "tool not detected" |
| lint | PASS/FAIL/SKIP | Warning/error count |
| test | PASS/FAIL/SKIP | N passed, M failed |
| build | PASS/FAIL/SKIP | Success or error summary |
| circular | PASS/WARN/SKIP | Circular dependency count (madge) |
| deadcode | INFO/SKIP | Unused export/file count (knip) |
| secrets | PASS/CRITICAL/SKIP | Detected secret count (gitleaks) |
| ai-slop | PASS/WARN/SKIP | console.log/any count (ast-grep) |
Verdict: PASS = exit code 0, FAIL = 3 retries exhausted, SKIP = tool not detected Overall verdict: Any CRITICAL → FAIL. All PASS/WARN/INFO/SKIP → PASS. Any FAIL among typecheck/lint/test/build → FAIL.
Fix Strategy by Error Category
| Category | Priority | Fix Method |
|---|---|---|
| TYPE | High | Extract file:line from output → Read the file → fix the type |
| LINT | Medium | --fix auto-fix first → manual fix |
| TEST | High | Analyze failing tests → fix the code (modifying tests is a last resort) |
| BUILD | High | Analyze build logs → check import/export issues first |
Max Retries
Up to 3 attempts per step. After 3 failures, stop and report to the user.