Commit Validator

References (archive): SCAFFOLD_SKILLS_ARCHIVE_MAP.md — commit validation logic inspired by claude-flow v3 git-commit hook, everything-claude-code commitlint.

Step 1: Validate Commit Message

Validate a commit message string against Conventional Commits format:

Format: <type>(<scope>): <subject>

Types:

• feat: A new feature
• fix: A bug fix
• docs: Documentation only changes
• style: Code style changes (formatting, etc.)
• refactor: Code refactoring
• perf: Performance improvements
• test: Adding or updating tests
• chore: Maintenance tasks
• ci: CI/CD changes
• build: Build system changes
• revert: Reverting a previous commit

Validation Rules:

1. Must start with type (required)
2. Scope is optional (in parentheses)
3. Subject is required (after colon and space)
4. Use imperative, present tense ("add" not "added")
5. Don't capitalize first letter
6. No period at end
7. Can include body and footer (separated by blank line)

Use this regex pattern for validation:

const CONVENTIONAL_COMMIT_REGEX =
  /^(feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert)(\(.+\))?: .{1,72}/;

function validateCommitMessage(message) {
  const lines = message.trim().split('\n');
  const header = lines[0];

  // Check format
  if (!CONVENTIONAL_COMMIT_REGEX.test(header)) {
    return {
      valid: false,
      error: 'Commit message does not follow Conventional Commits format',
    };
  }

  // Check length
  if (header.length > 72) {
    return {
      valid: false,
      error: 'Commit header exceeds 72 characters',
    };
  }

  return { valid: true };
}

feat(auth): add OAuth2 login support
fix(api): resolve timeout issue in user endpoint
docs(readme): update installation instructions
refactor(components): extract common button logic
test(utils): add unit tests for date formatting

Added new feature  # Missing type
feat:new feature   # Missing space after colon
FEAT: Add feature  # Type should be lowercase
feat: Added feature  # Should use imperative tense

#!/bin/bash
commit_msg=$(git log -1 --pretty=%B)
if ! node .claude/tools/cli/validate-commit.mjs "$commit_msg"; then
  echo "Commit message validation failed"
  exit 1
fi

# .github/workflows/validate-commits.yml
- name: Validate commit messages
  run: |
    git log origin/main..HEAD --pretty=%B | while read msg; do
      node .claude/tools/cli/validate-commit.mjs "$msg" || exit 1
    done

Returns structured validation result:

{
  "valid": true,
  "type": "feat",
  "scope": "auth",
  "subject": "add OAuth2 login support",
  "warnings": []
}

Or for invalid messages:

{
  "valid": false,
  "error": "Commit message does not follow Conventional Commits format",
  "suggestions": [
    "Use format: <type>(<scope>): <subject>",
    "Valid types: feat, fix, docs, style, refactor, perf, test, chore, ci, build, revert"
  ]
}

# Validate a commit message
node .claude/tools/cli/validate-commit.mjs "feat(auth): implement jwt login"

# Validate from stdin (e.g. in a hook)
echo "fix: incorrect variable name" | node .claude/tools/cli/validate-commit.mjs

Iron Laws

1. ALWAYS validate commit messages in both pre-commit hook and CI — pre-commit catches local violations; CI catches cases where the hook was bypassed or not installed; both layers are required.
2. NEVER accept commit messages without a type prefix — conventional commit format (type: subject) is the foundation; messages without a type are unparseable for changelog generation and semantic versioning.
3. ALWAYS enforce subject line length limit (72 characters) — subjects over 72 characters are truncated in git log --oneline and GitHub PR views; conciseness is enforced, not just encouraged.
4. NEVER block commits for body/footer format issues — only type, subject, and length are blocking; optional sections (body, footer, co-authorship) should warn, not block.
5. ALWAYS provide the correct format example in rejection messages — error messages without examples cause developers to guess the format; show feat: add user authentication alongside every rejection.

Anti-Patterns

Memory Protocol (MANDATORY)

Before starting: Read .claude/context/memory/learnings.md

After completing:

• New pattern -> .claude/context/memory/learnings.md
• Issue found -> .claude/context/memory/issues.md
• Decision made -> .claude/context/memory/decisions.md

ASSUME INTERRUPTION: If it's not in memory, it didn't happen.