Are you an LLM? Read llms.txt for a summary of the docs, or llms-full.txt for the full context.
Skip to content

Validation Skill

Defines quality gates and tool detection procedures for production readiness checks.

What It Provides

Quality Gate Definitions
  • What to check (tests, types, lint, coverage, security)
  • Pass/fail criteria for each gate
  • Adaptation strategies for different tech stacks
Tool Detection Procedures
  • How to detect test frameworks (Vitest, Jest, Playwright)
  • How to detect type checkers (TypeScript, Flow)
  • How to detect linters (ESLint, Biome)
  • How to detect coverage tools
Validation Execution Methods
  • Command sequences to run for each tool
  • Output parsing patterns
  • Error reporting formats

Quality Gates

Tests

What to check: All tests passing

Detection: 
# Check package.json scripts
- npm test → run test command
- check for: vitest, jest, @playwright/test

Pass criteria: Exit code 0, no failures

Adaptation:
  • No test script → Skip, warn user
  • Multiple test types → Run all (unit, integration, e2e)

Type Checking

What to check: No type errors

Detection: 
# Check for tsconfig.json
if exists → run: tsc --noEmit
 
# Check for Flow
if .flowconfig → run: flow check

Pass criteria: 0 errors

Adaptation:
  • JavaScript only → Skip type checking
  • @ts-check comments → Validate those files

Linting

What to check: 0 lint errors

Detection: 
# Check for ESLint config
.eslintrc.* or package.json "eslintConfig" → run: eslint .
 
# Check for Biome
biome.json → run: biome check .

Pass criteria: 0 errors (warnings OK)

Adaptation:
  • No linter → Skip

Dead Code

What to check: No unused exports/imports

Method: 
# Build dependency graph
1. Find all exports
2. Find all imports
3. Mark exports never imported as dead
4. Mark imports never used as dead

Pass criteria: 0 unused exports/imports

Note: Detection only. Removal delegated to fixer agent.

Coverage

What to check: ≥80% coverage

Detection: 
# Check test runner config
- Vitest → vitest.config.ts coverage settings
- Jest → jest.config.js coverage settings

Pass criteria: 80% average (statement, branch, function, line)

Adaptation:
  • No coverage config → Warn, don't fail
  • Custom threshold in package.json → Use that

Security

What to check: No high/critical vulnerabilities

Method: 
npm audit
# or
yarn audit
# or
pnpm audit

Pass criteria: 0 high or critical vulnerabilities

How Agents Use This

Reviewer Agent:
  1. Reads this skill for validation procedures
  2. Executes each quality gate check
  3. Reports results
  4. Delegates fixes to fixer agent if needed
Fixer Agent:
  1. Reads validation results
  2. Applies auto-fixes where possible
  3. Triggers re-validation

Validation Output Format

{
  "gates": {
    "tests": {
      "passed": true,
      "details": "47 tests passed"
    },
    "types": {
      "passed": false,
      "errors": [
        {
          "file": "src/components/LoginForm.tsx",
          "line": 23,
          "message": "Type 'string' not assignable to type 'number'"
        }
      ]
    },
    "lint": {
      "passed": true,
      "details": "0 errors"
    },
    "coverage": {
      "passed": false,
      "percentage": 76,
      "threshold": 80,
      "uncovered": [
        "src/components/TaskItem.tsx",
        "lib/utils.ts"
      ]
    },
    "security": {
      "passed": true,
      "vulnerabilities": 0
    }
  },
  "overall_passed": false
}

Auto-Fix Capabilities

This skill defines what's auto-fixable:

Auto-fixable:

  • Unused imports → Remove
  • Lint errors with --fix → Apply fixes
  • Simple type errors → Add annotations

Not auto-fixable:

  • Test failures → Logic bugs
  • Complex type errors → Design issues
  • Coverage gaps → Need new tests
  • Security vulnerabilities → Need updates

Configuration

Override defaults in package.json:

{
  "agentful": {
    "validation": {
      "coverageThreshold": 90,
      "typeCheck": false,
      "testCommand": "vitest run --coverage",
      "e2eCommand": "playwright test"
    }
  }
}

Implementation Details

Location: .claude/skills/validation/SKILL.md

Model: Sonnet

Tools Used:
  • Read (config files)
  • Glob (find test files)
  • Grep (search patterns)
  • Bash (run validation commands)
Config Files Read:
  • package.json
  • tsconfig.json
  • .eslintrc.*
  • biome.json
  • vitest.config.ts / jest.config.js
  • playwright.config.ts / cypress.config.ts

Error Reporting Patterns

Test Failure

❌ Test failed: User authentication › should log in with valid credentials
   Expected: true
   Received: false
   at src/__tests__/auth.test.ts:45

Type Error

❌ Type error in LoginForm.tsx:23
   Type 'string' is not assignable to type 'number'

Coverage Gap

⚠️  Coverage below threshold: 76% (target: 80%)
   Missing coverage:
   • src/components/TaskItem.tsx (65%)
   • lib/utils.ts (50%)

Usage by Commands

/agentful-validate
  • Primary consumer of this skill
  • Runs all quality gate checks
  • Reports results to user
Orchestrator Agent
  • Uses validation before marking features complete
  • Delegates to reviewer agent which uses this skill

Extending the Skill

Add custom quality gates by editing .claude/skills/validation/SKILL.md:

## Custom Gate: API Contract Validation
 
What to check: OpenAPI spec matches implementation
 
Detection:
- Find openapi.yml or swagger.json
- Run: npm run validate:api-contract
 
Pass criteria: 0 mismatches
 
Command:
\`\`\`bash
npx swagger-cli validate openapi.yml
\`\`\`

See Also