feature-validate
Feature Validate Skill
Run comprehensive quality gates on the implemented feature. Phase 4 of the feature lifecycle (design → plan → implement → validate → release).
Instructions
Phase 0: PRIME
- Read and follow the repository CLAUDE.md before any other action — it may override defaults for linting, test commands, or gate criteria.
- Verify feature state is
validateandimplementis completed. All state operations go throughpython3 ~/.claude/scripts/feature-state.py— never modify state files directly. - Load implementation artifact from
.feature/state/implement/. Validation cannot proceed without implementation artifacts; if missing, stop and report. - Load L0, L1, and implementation context so quality gates run against the correct scope.
Gate: Implementation artifact loaded. Feature state confirms implement completed. Proceed.
Phase 1: EXECUTE (Quality Gates)
Step 1: Language Detection
Auto-detect project language(s) from file extensions, build files, and the implementation artifact. This detection drives which quality gate skill runs next.
Step 2: Run Quality Gates
Use the repository's existing quality gate skills — do not re-implement linting or test runners inline. Route to the appropriate skill per language:
| Language | Quality Gate | Command |
|---|---|---|
| Go | go-pr-quality-gate | make check or go test ./... && go vet ./... |
| Python | python-quality-gate | ruff check && pytest && mypy |
| TypeScript | universal-quality-gate | npm run typecheck && npm run lint && npm test |
| Other | universal-quality-gate | Detect and run project-specific checks |
Show the full, unedited command output for every gate — never summarize or truncate test results, because summaries hide the exact failure details needed for diagnosis.
Step 3: Regression Check
Run the full test suite and compare against pre-implementation baseline:
- New test failures = regression
- Missing tests for new code = coverage gap
Existing tests must still pass; new failures here block advancement regardless of new-feature test results.
Step 4: Custom Gates
If the design document specified custom validation criteria, check those too.
Step 5 (optional): Security and Performance
If explicitly enabled by the user:
- Security scan: Run security-focused review agent
- Performance check: Run benchmarks against baseline
These are off by default — do not run them unless the user requests it.
Gate: Every gate must pass. No gate may be skipped and no failure may be approved — a single failing gate blocks advancement to release. Proceed only when all results are green.
Phase 2: VALIDATE (Report)
Produce the validation report:
# Validation Report: [Feature Name]
## Quality Gates
- [ ] Tests: PASS/FAIL (X/Y passed)
- [ ] Lint: PASS/FAIL (N issues)
- [ ] Type Check: PASS/FAIL
- [ ] Regression: PASS/FAIL
## Coverage
- New files: X
- Modified files: Y
- Test coverage: Z%
## Issues Found
- [Issue 1]: [severity]
- [Issue 2]: [severity]
## Verdict: PASS / NEEDS_FIXES / BLOCK
The verdict must reflect actual gate results — never mark PASS if any gate failed.
If NEEDS_FIXES: suggest running /feature-implement with specific fix tasks. This skill does not fix failing tests; it reports them and routes back.
If BLOCK: explain blocking issues.
Gate: Report produced. Proceed to Checkpoint.
Phase 3: CHECKPOINT
-
Save validation artifact (all state operations through the feature-state script):
echo "VALIDATION_REPORT" | python3 ~/.claude/scripts/feature-state.py checkpoint FEATURE validate -
Record learnings — if this phase produced non-obvious insights, record them:
python3 ~/.claude/scripts/learning-db.py record TOPIC KEY "VALUE" --category design -
If validation passed, advance:
python3 ~/.claude/scripts/feature-state.py advance FEATURE -
Suggest next step:
Validation passed. Run /feature-release to merge and release.
Error Handling
| Error | Cause | Solution |
|---|---|---|
| Tests fail | Implementation bugs | Route back to /feature-implement with failure details |
| No test framework detected | Project setup incomplete | Report gap, suggest setup |
| Lint failures | Style issues | Auto-fix if trivial, otherwise route back |
References
More from notque/claude-code-toolkit
generate-claudemd
Generate project-specific CLAUDE.md from repo analysis.
12fish-shell-config
Fish shell configuration and PATH management.
12pptx-generator
PPTX presentation generation with visual QA: slides, pitch decks.
12codebase-overview
Systematic codebase exploration and architecture mapping.
10image-to-video
FFmpeg-based video creation from image and audio.
9data-analysis
Decision-first data analysis with statistical rigor gates.
9