quality-gates（质量门）

Quality Gates Skill

Objective

Enforce code quality standards by running automated checks that must pass before code can be committed, merged, or deployed. Acts as a guardian ensuring consistent quality across the codebase.

When to Use This Skill

Auto-invoke when:

• User completes feature implementation
• Before creating commits or pull requests
• User asks to "test", "validate", "check quality", or "verify"
• Before deployment or release
• After significant refactoring

Quality Gate Levels

Level 1: Pre-Commit Gates (Fast, < 30 seconds)

Essential checks that run before every commit.

Level 2: Pre-Push Gates (Moderate, < 2 minutes)

Comprehensive checks before pushing to remote.

Level 3: Pre-Deploy Gates (Thorough, < 5 minutes)

Complete validation before production deployment.

Gate Execution Workflow

Gate 1: Linting (JavaScript/TypeScript)

Purpose: Enforce code style and catch common errors

Tools: Bash, Read

Process:

1. Detect linter by checking for:

- ESLint: .eslintrc*, eslint.config.* - Biome: biome.json - None: Skip this gate

1. Read package.json to find lint script: "scripts": {"lint": "eslint.", "lint:fix": "eslint. --fix"}
2. Execute linter: # Try to run lint script npm run lint # If fails, try direct commands npx eslint. || npx biome check.
3. Parse results:

- Exit code 0: ✅ PASS - Exit code non-zero: ❌ FAIL - Extract error count and file locations

1. Auto-fix attempt (if failures found): npm run lint:fix || npx eslint. --fix

Success Criteria: Zero linting errors (warnings acceptable)

Gate 2: Type Checking (TypeScript)

Purpose: Verify type safety and catch type errors

Tools: Bash, Read, Grep

Process:

1. Detect TypeScript by checking for:

- tsconfig.json - TypeScript in dependencies

1. Read tsconfig.json to check strictness:

- strict: true - noImplicitAny, strictNullChecks, etc.

1. Execute type checker: # Try to run typecheck script npm run typecheck || npm run type-check # If no script, run directly npx tsc --noEmit
2. Parse results:

- Exit code 0: ✅ PASS - Exit code non-zero: ❌ FAIL - Extract error count and locations

Success Criteria: Zero type errors

Gate 3: Unit & Integration Tests

Purpose: Verify code functionality and prevent regressions

Tools: Bash, Read, Grep

Process:

1. Detect test framework:

- Vitest: vitest.config.*, vitest in dependencies - Jest: jest.config.*, jest in dependencies - Native test: --test flag with Node.js 20+

1. Count test files: # Use Grep to find test files find. -name "*.test.*" -o -name "*.spec.*" | wc -l
2. Execute tests: # Run unit tests (fast) npm run test || npm run test:unit # Or direct command npx vitest run || npx jest --ci
3. Parse results:

- Total tests run - Passed / Failed / Skipped - Coverage percentage (if available)

1. Coverage check (if configured): npm run test:coverage # Check if meets threshold (e.g., 80%)

Success Criteria:

• All tests pass (100%)
• Coverage ≥ configured threshold (if set)

Gate 4: Build Verification

Purpose: Ensure code compiles and builds without errors

Tools: Bash

Process:

1. Detect build system:

- Next.js: next build - Vite: vite build - Webpack: webpack --mode production - TypeScript: tsc

1. Execute build: npm run build
2. Check build artifacts:

- Verify output directory exists: dist/, build/, .next/ - Check for build errors in logs

1. Clean up (optional): # Remove build artifacts to save space rm -rf dist/ build/.next/

Success Criteria: Build completes with exit code 0

Gate 5: Security Audit

Purpose: Identify known vulnerabilities in dependencies

Tools: Bash, Read

Process:

1. Run npm/pnpm audit: npm audit --json || pnpm audit --json
2. Parse audit results:

- Critical vulnerabilities: 0 - High vulnerabilities: 0 - Moderate vulnerabilities: < threshold - Low vulnerabilities: informational

1. Check for specific vulnerabilities:

- Prototype pollution - Remote code execution (RCE) - SQL injection - Cross-site scripting (XSS)

1. Suggest fixes: npm audit fix # or npm audit fix --force # (if safe)

Success Criteria:

• Zero critical/high vulnerabilities
• Moderate vulnerabilities acknowledged or fixed

Gate 6: Code Complexity Analysis (Optional)

Purpose: Flag overly complex code that may need refactoring

Tools: Grep, Bash

Process:

1. Detect code complexity tools:

- eslint-plugin-complexity - SonarQube - CodeClimate

1. Basic complexity checks: # Find files with excessive lines find src -name "*.{ts,tsx,js,jsx}" -exec wc -l {} \; | awk '$1 > 500' # Find deeply nested code (>5 levels) grep -rn "^[[:space:]]\{20,\}" src/ # Count TODO/FIXME grep -rn "TODO\|FIXME\|HACK" src/ | wc -l

Success Criteria:

• No files > 500 lines (warning only)
• No nesting > 5 levels (warning only)

Gate 7: Git Pre-Commit Checks

Purpose: Ensure commit quality and prevent sensitive data leaks

Tools: Bash, Grep

Process:

1. Check for sensitive data: # Search for API keys, secrets, tokens git diff --cached | grep -i "api[_-]key\|secret\|password\|token" # Check for.env files being committed git diff --cached --name-only | grep "\.env$"
2. Validate commit message (if Conventional Commits):

- Format: type(scope): description - Types: feat, fix, docs, style, refactor, test, chore

1. Check file sizes: # Flag files > 1MB git diff --cached --name-only | xargs ls -lh | awk '$5 > 1000000'

Success Criteria:

• No secrets in diff
• No.env files
• No large files (> 1MB)

Execution Strategy

Sequential Execution (Default)

Run gates in order, stop on first failure:

Lint → TypeCheck → Test → Build → Audit

Parallel Execution (Fast Mode)

Run independent gates simultaneously:

[Lint + TypeCheck + Test] → Build → Audit

Selective Execution

Run only relevant gates based on changes:

• .ts/.tsx files changed → TypeCheck
• Dependencies updated → Audit
• Test files changed → Tests only

Output Format

# Quality Gate Results

## Summary
✅ 5/7 Gates Passed | ❌ 2/7 Gates Failed

## Gate Details

### ✅ Gate 1: Linting
- **Status**: PASS
- **Duration**: 3.2s
- **Details**: 0 errors, 2 warnings

### ❌ Gate 2: Type Checking
- **Status**: FAIL
- **Duration**: 5.1s
- **Errors**: 3 type errors found
  - `src/components/Button.tsx:15` - Property 'onClick' is missing
  - `src/utils/api.ts:42` - Type 'string' is not assignable to type 'number'
  - `src/hooks/useAuth.ts:8` - Cannot find name 'User'

### ✅ Gate 3: Tests
- **Status**: PASS
- **Duration**: 12.4s
- **Tests**: 124 passed, 0 failed, 2 skipped
- **Coverage**: 87% (target: 80%)

### ⏭️ Gate 4: Build
- **Status**: SKIPPED (previous gate failed)

### ⏭️ Gate 5: Security Audit
- **Status**: SKIPPED (previous gate failed)

## Action Required
Fix the 3 type errors in Gate 2 before proceeding.

## Recommendations
1. Run `npm run typecheck` locally to see full error details
2. Consider adding pre-commit hooks to catch these earlier
3. Current code coverage (87%) exceeds target - excellent work!

Integration with Git Hooks

Setup Husky + lint-staged (Recommended)

Check if installed:

test -d .husky && echo "Husky installed" || echo "Husky not found"

Suggest installation if missing:

npm install --save-dev husky lint-staged
npx husky init

Configure.husky/pre-commit:

#!/usr/bin/env sh
. "$(dirname -- "$0")/_/husky.sh"

# Run quality gates
npm run lint
npm run typecheck
npm run test

Alternative: git commit -m with manual checks

If no hooks, prompt user:

⚠️  No pre-commit hooks detected.
Would you like me to run quality gates before committing? (Recommended)

Progressive Quality Gates

Level 1: Essential (Always Run)

• Linting
• Type checking

Level 2: Standard (Pre-Push)

• Essential +
• Unit tests
• Security audit

Level 3: Comprehensive (Pre-Deploy)

• Standard +
• Integration tests
• E2E tests
• Build verification
• Performance tests

Error Recovery

Auto-Fix Capability

• Lint errors: Run eslint --fix or biome check --apply
• Format errors: Run prettier --write
• Security vulnerabilities: Run npm audit fix

Manual Fix Required

• Type errors
• Test failures
• Build errors

Bypass (Use with Caution)

# Skip hooks for emergency fixes only
git commit --no-verify -m "emergency: fix critical bug"

Best Practices

1. Fail Fast: Stop at first critical failure to save time
2. Clear Feedback: Always show which gate failed and why
3. Actionable: Provide exact commands to fix issues
4. Configurable: Respect project's quality thresholds
5. Performance: Cache results when possible
6. Incremental: Only check changed files when appropriate

Configuration

Read from package.json

{
  "qualityGates": {
    "coverage": {
      "minimum": 80,
      "enabled": true
    },
    "audit": {
      "level": "moderate",
      "enabled": true
    },
    "complexity": {
      "maxLines": 500,
      "maxDepth": 5
    }
  }
}

Default Settings

If no config found, use sensible defaults:

• Coverage minimum: 70%
• Audit level: high/critical only
• Max file lines: 500
• Max nesting: 5 levels

Integration with Other Skills

• codebase-analysis - Use to detect available quality tools
• git-workflow - Integrate with commit/push process
• ci-cd-setup - Configure gates for CI pipeline

Version History

• 1.0.0 (2025-01-03): Initial skill with 7 quality gates and progressive execution