commit-message-generator

---
name: commit-message-generator
description: Automatically generate conventional commit messages when user has staged changes and mentions committing. Analyzes git diff and status to create properly formatted commit messages following conventional commits specification. AI attribution comes from the tool's own trailer (Claude Code's Co-Authored-By); the skill never adds a second one. Invoke when user mentions "commit", "staged", "committing", or asks for help with commit messages.
---

# Commit Message Generator

Automatically generate conventional commit messages for staged changes.

## Philosophy

Commit messages are the history and documentation of your code's evolution.

### Core Beliefs

1. **Commits Tell a Story**: Each commit should explain what changed and why
2. **Conventional Format Enables Automation**: Structured messages power changelogs and semantic versioning
3. **Clarity Over Brevity**: A clear 2-line message beats a cryptic 5-word one
4. **Context Matters**: Messages should be understandable months later without additional context

### Why Conventional Commits

- **Automated Changelogs**: Generate release notes from commit history
- **Semantic Versioning**: Determine version bumps (major/minor/patch) automatically
- **Better Searchability**: Find specific types of changes quickly
- **Team Communication**: Consistent format aids code review and collaboration

## When to Use This Skill

Activate this skill when the user:
- Mentions "commit", "committing", "staged changes", or "ready to commit"
- Shows `git add` or `git status` output with staged changes
- Asks "what should my commit message be?"
- Says "I need to commit my changes"
- Asks for help writing commit messages

## Decision Framework

Before generating a commit message, ask yourself:

### Is This Ready to Commit?

- ✅ **Yes** - Staged changes represent a single logical unit of work
- ❌ **No** - Multiple unrelated changes staged → Suggest splitting into separate commits
- ⚠️ **Maybe** - Large changeset → Review to ensure it's cohesive

### What Type of Change Is This?

1. **New functionality** → `feat` type
2. **Bug fix** → `fix` type
3. **Code improvement without behavior change** → `refactor` type
4. **Documentation only** → `docs` type
5. **Tests only** → `test` type
6. **Multiple types** → Suggest splitting commits

### What Scope Makes Sense?

- **Module/component name** - For focused changes (e.g., `auth`, `api`, `ui`)
- **Feature area** - For cross-cutting changes (e.g., `validation`, `logging`)
- **No scope** - For global changes (e.g., dependencies, config)

### Should This Be Multiple Commits?

Split if staged changes include:
- ✅ Unrelated features or fixes
- ✅ Refactoring + new feature (split: refactor first, feature second)
- ✅ Multiple bug fixes
- ❌ Feature + tests (keep together)
- ❌ Feature + documentation (keep together)

### Decision Tree

```
User mentions commit
    ↓
Check: Staged changes?
    ↓ Yes
Check: Multiple unrelated changes?
    ↓ No
Check: Follows conventional commits pattern?
    ↓ Generate message
    ↓
Review with user → Commit
```

## Workflow

### 1. Check for Staged Changes

```bash
git status
git diff --staged
```

If no staged changes, inform the user and suggest staging files first.

### 2. Analyze Recent Commits for Style

```bash
git log --oneline -10
```

Learn the repository's commit message conventions.

### 3. Generate Conventional Commit Message

**Format**: `<type>(<scope>): <description>`

**Types**:
- `feat` - New feature
- `fix` - Bug fix
- `docs` - Documentation changes
- `style` - Code style changes (formatting, semicolons, etc.)
- `refactor` - Code refactoring (no functional changes)
- `test` - Adding or updating tests
- `chore` - Build process, dependency updates, etc.
- `perf` - Performance improvements
- `ci` - CI/CD changes

**Example**:
```
feat(auth): add two-factor authentication support

- Implement TOTP-based 2FA
- Add backup codes generation
- Include recovery flow for lost devices
- Update user profile settings UI

Refs: 30024167
```

**Body rules**:

- **Wrap the body at 72 columns**, per git convention. This is the opposite of
  PR and comment bodies, which must never be hard-wrapped (`pr-create`) —
  don't carry one rule into the other.
- **Take the ticket from the branch name into a `Refs:` footer.** Kanopi
  branches are `<type>/<ticketID>-<short-desc>`, so
  `bug/30140451-menu-link-attributes` gives `Refs: 30140451`. Omit the footer
  when the branch carries no ticket.

### 4. Drupal/WordPress-Specific Patterns

**Drupal**:
- Config changes: `feat(config): add user profile field configuration`
- Module work: `fix(custom_module): correct permission check in access callback`
- Hooks: `refactor(hooks): simplify hook_form_alter implementation`

**WordPress**:
- Theme work: `style(theme): improve mobile navigation styles`
- Plugin work: `fix(plugin): correct ACF field validation`
- Blocks: `feat(blocks): add testimonial Gutenberg block`

### 5. Leave AI Attribution to the Tool

Claude Code adds its own AI attribution trailer to commits it creates — by default
`Co-Authored-By: Claude <model> <noreply@anthropic.com>`, carrying the session's
real model, and governed by the `attribution.commit` setting. That trailer is the
per-commit provenance record. The per-PR complement is the Delivery Record
([kanopi/delivery-record](https://github.com/kanopi/delivery-record)).

Rules:

- **Add no attribution trailer of your own** — no `Assisted-by:`, and no second
  `Co-Authored-By:`. Two trailers for one assistant split the audit trail.
- **Never edit or remove the tool's trailer**, and never type a model name into
  it yourself; the tool supplies it.
- **Keep the trailers in one final block.** Git reads trailers only from the
  last paragraph, so put `Refs:` and `BREAKING CHANGE:` footers directly above
  the tool's trailer, with no blank line between them and nothing after.
- **If the session's attribution settings say to add none**, add none.
- **A human-written change** still gets Claude Code's trailer on any commit
  Claude Code makes. If AI only phrased the message and the trailer would
  misstate who wrote the change, have the human run the commit.

The audit trail:

```bash
git log --grep="^Co-Authored-By: Claude"   # every Claude Code commit
git log --grep="^Assisted-by:"             # commits from before this convention
```

### 6. Present to User for Approval

Show the generated commit message in a clear code block and ask:

> "Here's a commit message based on your staged changes. Would you like me to commit with this message, or would you like to modify it?"

Wait for explicit user approval (e.g., "approve", "yes, commit", or an edited version) before running `git commit`.

### 7. Execute Commit (only after approval)

```bash
git commit -m "commit message here"
```

Or for multi-line messages:
```bash
git commit -m "$(cat <<'EOF'
feat(auth): add two-factor authentication support

- Implement TOTP-based 2FA
- Add backup codes generation
- Include recovery flow

Refs: PROJ-123
Co-Authored-By: Claude <model> <noreply@anthropic.com>
EOF
)"
```

## Integration with CMS Cultivator

This skill is invoked directly by the main session when the user mentions committing — there is no orchestrator agent in between. The skill works the same way whether triggered conversationally ("I'm ready to commit") or explicitly ("generate a commit message").

The companion **`pr-create`** skill picks up where this one ends — once you've committed, ask to "create a PR" and `pr-create` will generate the PR description from your commits. The tool's attribution trailer — Claude Code's `Co-Authored-By:` — is the per-commit half of Kanopi's provenance model; the per-PR half is the Delivery Record ([kanopi/delivery-record](https://github.com/kanopi/delivery-record)).

## Best Practices

1. **Analyze the actual changes** - Don't just use file names, understand what changed
2. **Be specific in scope** - Use module name, component name, or feature area
3. **Include breaking changes** - Add `BREAKING CHANGE:` footer if applicable
4. **Reference tickets** - Include ticket numbers if found in branch name
5. **Follow project conventions** - Match the existing commit style

## Example Interactions

**Example 1: Simple fix**
```
User: "I fixed the login bug, ready to commit"
[Skill analyzes git diff]
Skill: "I'll generate a commit message for your login fix..."

Generated:
fix(auth): resolve session timeout on remember-me login

- Correct cookie expiration logic
- Add test coverage for remember-me flow

Co-Authored-By: Claude <model> <noreply@anthropic.com>

Would you like me to commit with this message?
```

**Example 2: Feature addition**
```
User: "Staged my changes, need to commit"
[Skill analyzes extensive changes]
Skill: "I see you've added a new dashboard widget. Here's a commit message..."

Generated:
feat(dashboard): add performance metrics widget

- Create new widget component
- Add database query for metrics
- Include responsive design
- Add widget to admin dashboard page
```

## Resources

- [Conventional Commits](https://www.conventionalcommits.org/)
- [Angular Commit Guidelines](https://github.com/angular/angular/blob/master/CONTRIBUTING.md#commit)
- [Semantic Commit Messages](https://gist.github.com/joshbuchea/6f47e86d2510bce28f8e7f42ae84c716)