AgentSkillsCN

Code Verify

代码验证

SKILL.md

Code Verify Skill

Code verification and evidence logging for quality gates and governance compliance.

Philosophy

"Quality is not an act, it is a habit." - Aristotle "Evidence of quality is not optional for enterprise software."

Code verification provides auditable proof that quality gates passed at key development milestones.

When to Activate

This skill should activate when:

  • User mentions "verify", "check", "validate" in code context
  • A feature or section is completed
  • User is about to commit changes
  • User asks about code quality or testing
  • Discussion of governance requirements or evidence

Trigger Phrases

Immediate Verification Triggers

PhraseAction
"verify this"Run full verification
"verify that"Run full verification
"check this"Run full verification
"check the code"Run full verification
"run checks"Run full verification
"run verification"Run full verification
"validate code"Run full verification
"validate this"Run full verification

Quick Check Triggers

PhraseAction
"quick check"Run analysis only
"quick verify"Run analysis only
"lint check"Run analysis only
"analyze this"Run analysis only

Pre-commit Triggers

PhraseAction
"pre-commit check"Run analysis + tests
"before commit"Run analysis + tests
"ready to commit"Suggest pre-commit check
"about to commit"Suggest pre-commit check

Section/Feature Complete Triggers

PhraseAction
"section complete"Prompt for verification
"feature done"Prompt for verification
"finished the"Prompt for verification
"completed the"Prompt for verification

Verification Levels

Full Verification

All checks with evidence logging:

  1. Static analysis (lint, type check)
  2. Format verification
  3. Unit tests
  4. Build verification

Quick Verification

Fast feedback during development:

  1. Static analysis
  2. Format check

Pre-commit Verification

Balance of speed and safety:

  1. Static analysis
  2. Format check
  3. Tests

Technology-Specific Commands

Flutter/Dart

bash
flutter analyze --fatal-infos --fatal-warnings
dart format --set-exit-if-changed .
flutter test
flutter build web

TypeScript/React

bash
npm run lint
npm run type-check
npm run format:check
npm run test
npm run build

Python

bash
ruff check .
mypy .
black --check .
pytest

Node.js

bash
npm run lint
npm run test
npm run build

Evidence Logging

All verification runs produce:

Verification Report

planning/verification-logs/YYYY-MM-DD_HHMM_[feature].md

markdown
# Verification Report: {feature}

**Date**: {date}
**Time**: {time}
**Trigger**: {what triggered verification}

## Summary

| Check | Status | Details |
|-------|--------|---------|
| Analysis | ✅/❌ | {count} issues |
| Format | ✅/❌ | {changed files} |
| Tests | ✅/❌ | {pass}/{total} |
| Build | ✅/❌ | {build status} |

## Details

### Static Analysis
{analysis output}

### Tests
{test output summary}

### Build
{build output summary}

## Files in Scope
- {file1}
- {file2}

Raw Log

planning/verification-logs/raw/YYYY-MM-DD_HHMM_[feature].log

  • Complete stdout/stderr
  • Timestamps for each step
  • Exit codes

Index Entry

planning/verification-logs/index.md

  • Running table of all verification runs
  • Quick reference for auditors

Integration Points

With Feature Workflow

When a feature is completed, prompt:

"Feature complete! Should I run verification and log evidence?"

With Commits

When about to commit, suggest:

"Run pre-commit verification first?"

With Scribe

At session end, if code was written:

"Run verification before wrapping up?"

With Stage Review

Pre/post build gates include verification:

  • Pre-build: Baseline verification
  • Post-build: Full verification with evidence

Graceful Degradation

If tooling is not set up:

  1. Note which checks couldn't run
  2. Log what was attempted
  3. Still create evidence record
  4. Suggest setup steps
markdown
## Verification Attempted

| Check | Status | Note |
|-------|--------|------|
| Analysis | ⚠️ Skipped | Flutter not installed |
| Format | ⚠️ Skipped | Flutter not installed |
| Tests | ⚠️ Skipped | Flutter not installed |
| Build | ⚠️ Skipped | Flutter not installed |

**Setup Required**: Run `flutter doctor` to configure Flutter

Example Flow

code
User: "I've finished the user profile widget, verify this"

Claude:
1. Detects "verify this" trigger
2. Identifies context: user profile widget
3. Runs full verification:
   - flutter analyze → ✅ 0 issues
   - dart format → ✅ no changes
   - flutter test → ✅ 24/24
   - flutter build → ✅ compiled
4. Generates report
5. Updates index
6. Reports summary to user

Response:
"## Verification Complete ✅

Feature: user-profile-widget

| Check | Status |
|-------|--------|
| Analysis | ✅ Clean |
| Format | ✅ Clean |
| Tests | ✅ 24/24 |
| Build | ✅ OK |

📄 Evidence logged: planning/verification-logs/2026-01-05_1430_user-profile-widget.md"

Gentle Prompts

At natural breakpoints, consider suggesting:

"That feature looks complete. Want me to run verification and log evidence?"

But don't be pushy - verification is user-triggered by design.

Why This Matters

For enterprise/governance sign-off, you need to demonstrate:

  1. Systematic quality - Not random testing, structured verification
  2. Evidence trail - Timestamped proof, not just claims
  3. Tool-based validation - Automated analysis, not just eyeballing
  4. Natural integration - Quality as part of workflow, not afterthought

This skill ensures verification is easy to do and always documented.

Related Files

  • library/skills/verification-loop/ - The verification pattern itself
  • library/patterns/verification-loop.md - Documentation
  • planning/verification-logs/ - Evidence storage