メニュー
Best practices for creating high-quality Git commits. Covers commit message conventions (Conventional Commits), atomic commits, staging strategies, and when to amend vs create new commits. Helps AI agents generate clean, meaningful commit history.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Collaboration workflows for team-based Git development. Covers pull request workflows, merge vs rebase strategies, conflict resolution, code review practices, and branch protection. Helps AI agents follow team conventions and collaborate effectively.
Safety practices and recovery techniques for Git. Covers reflog usage, reset vs revert strategies, recovering lost commits, undoing mistakes, and preventing data loss. Helps AI agents safely navigate Git operations and recover from errors.
Modern Git command best practices for AI agents. Use modern, purposeful commands introduced in Git 2.23+ instead of legacy multi-purpose commands. Teaches when to use `git switch` (branch operations), `git restore` (file operations), and other safer alternatives to improve clarity and reduce errors.
| name | git-commits |
| description | Best practices for creating high-quality Git commits. Covers commit message conventions (Conventional Commits), atomic commits, staging strategies, and when to amend vs create new commits. Helps AI agents generate clean, meaningful commit history. |
Purpose: This skill teaches AI agents to create high-quality commits with clear messages, proper granularity, and effective use of the staging area.
<type>(<scope>): <subject>
<body>
<footer>
# ✓ CORRECT: Use these standard types
feat: # New feature for the user
fix: # Bug fix
docs: # Documentation changes
style: # Formatting, missing semicolons, etc. (no code change)
refactor: # Code change that neither fixes a bug nor adds a feature
perf: # Performance improvement
test: # Adding or updating tests
chore: # Maintenance tasks, dependencies, config
# Examples:
feat(auth): add password reset functionality
fix(api): handle null response in user endpoint
docs(readme): update installation instructions
refactor(parser): simplify token extraction logic
# ✓ GOOD: Clear, concise, imperative mood
feat(auth): add OAuth2 login support
fix(cart): prevent duplicate items
docs(api): document rate limiting
# ✗ BAD: Vague, past tense, too long
feat(auth): added some authentication stuff
fix(cart): fixed a bug
docs(api): updated the documentation to include information about rate limiting
Rules:
# ✓ GOOD: Explains WHY and context
feat(cache): add Redis caching layer
Improves API response time by 80% for frequently accessed data.
Uses Redis with 1-hour TTL for user profile and product catalog
endpoints. Falls back to database if Redis is unavailable.
Related to performance optimization initiative.
# ✗ BAD: Just repeats the subject
feat(cache): add Redis caching layer
Added Redis caching.
When to include body:
When to skip body:
# Breaking changes
feat(api): change user endpoint response format
BREAKING CHANGE: User API now returns camelCase instead of snake_case.
Migration guide: https://docs.example.com/migration/v2
# Issue references
fix(auth): prevent token expiration race condition
Closes #123
Refs #456
# Co-authors
feat(search): implement full-text search
Co-authored-by: Jane Doe <jane@example.com>
Definition: One commit = one logical change
# ✓ GOOD: Atomic commits
git commit -m "feat(auth): add login endpoint"
git commit -m "feat(auth): add logout endpoint"
git commit -m "test(auth): add login tests"
# ✗ BAD: Non-atomic commit
git commit -m "feat(auth): add login, logout, password reset, and tests"
# ✓ CORRECT: Stage related changes only
# You made changes to auth.js, api.js, and tests.js
# Commit 1: Feature implementation
git add src/auth.js src/api.js
git commit -m "feat(auth): add token refresh mechanism"
# Commit 2: Tests
git add tests/auth.test.js
git commit -m "test(auth): add token refresh tests"
# ✗ WRONG: Stage everything together
git add .
git commit -m "feat(auth): add token refresh and tests"
Does this change have a single, clear purpose?
├─ YES → Can it be described in one sentence?
│ ├─ YES → One commit ✓
│ └─ NO → Multiple commits
│
└─ NO → Multiple commits
Examples:
├─ "Add login validation" → One commit ✓
├─ "Add login and signup" → Two commits
└─ "Add login, fix bug, update docs" → Three commits
# You changed multiple things in one file
# Only stage the lines related to current commit
git add -p src/app.js
# Interactive prompts:
# y - stage this hunk
# n - don't stage this hunk
# s - split into smaller hunks
# e - manually edit the hunk
# Then commit only staged changes
git commit -m "feat(app): add error handling"
# Stage and commit remaining changes separately
git add -p src/app.js
git commit -m "refactor(app): extract validation logic"
# ✓ WORKFLOW: Create multiple atomic commits from current changes
# Check what changed
git status
git diff
# Commit 1: Feature A
git add src/feature-a.js src/utils.js
git commit -m "feat(feature-a): implement feature A"
# Commit 2: Feature B
git add src/feature-b.js
git commit -m "feat(feature-b): implement feature B"
# Commit 3: Tests
git add tests/
git commit -m "test: add tests for features A and B"
# Use case: You forgot to include a file in the last commit
# ✓ CORRECT: Amend if commit is NOT pushed yet
git add forgotten-file.js
git commit --amend --no-edit
# OR update the commit message too
git commit --amend -m "feat(auth): add login endpoint and middleware"
# ✗ WRONG: Amend if commit is already pushed
# This rewrites history and causes problems for collaborators
# If already pushed, create new commit instead:
git add forgotten-file.js
git commit -m "feat(auth): add missing middleware file"
# 1. Check current state
git status
git diff
# 2. Stage related changes only
git add src/feature.js src/api.js
# 3. Review what will be committed
git diff --staged
# 4. Commit with conventional format
git commit -m "feat(feature): add user profile customization
Allows users to customize avatar, bio, and theme preferences.
Settings are persisted to user preferences API endpoint.
Closes #234"
# 5. Verify commit
git log -1 --stat
# You worked on multiple things, now need to commit separately
# 1. Check all changes
git status
# 2. First commit: Core feature
git add src/core.js
git commit -m "feat(core): add data validation layer"
# 3. Second commit: API integration
git add src/api.js
git commit -m "feat(api): integrate validation with endpoints"
# 4. Third commit: Tests
git add tests/
git commit -m "test(validation): add comprehensive validation tests"
# 5. Fourth commit: Documentation
git add docs/
git commit -m "docs(validation): document validation rules"
# ✓ GOOD: Clear bug fix commit
git add src/buggy-file.js
git commit -m "fix(cart): prevent duplicate items on double-click
Added debounce to 'Add to Cart' button and server-side duplicate
check. Prevents race condition when users click rapidly.
Fixes #567"
# Include:
# - What the bug was
# - How you fixed it
# - Issue reference
Did I already push this commit?
├─ NO → Safe to amend
│ └─> git commit --amend
│
└─ YES → Create new commit instead
└─> git commit -m "fix: ..."
Is this a fixup for recent commit?
└─> Consider git commit --fixup=<sha>
Then use git rebase -i --autosquash before pushing
# ✓ GOOD: Separate refactoring from behavior changes
# Commit 1: Pure refactor (no behavior change)
git add src/parser.js
git commit -m "refactor(parser): extract token validation to separate function
No behavior change. Makes code more testable and readable."
# Commit 2: Behavior change
git add src/parser.js
git commit -m "feat(parser): add support for nested tokens
Now supports tokens in format {{parent.child.value}}"
# Why separate?
# - Easy to verify refactor doesn't change behavior
# - Easy to revert feature without losing refactor
# - Clear history
# ✗ BAD: No context
git commit -m "update"
git commit -m "fix bug"
git commit -m "changes"
git commit -m "wip"
# ✓ GOOD: Clear and specific
git commit -m "feat(auth): add JWT token expiration check"
git commit -m "fix(api): handle null values in user profile endpoint"
git commit -m "refactor(parser): simplify regex patterns"
git commit -m "docs(readme): add Docker setup instructions"
# ✗ BAD: Multiple unrelated changes in one commit
git add .
git commit -m "feat: add login, fix cart bug, update docs"
# ✓ GOOD: Separate commits
git add src/auth.js
git commit -m "feat(auth): add OAuth login support"
git add src/cart.js
git commit -m "fix(cart): prevent negative quantities"
git add docs/
git commit -m "docs(api): update authentication endpoints"
# ✗ BAD: Leaving console.log, debugger statements
git add src/app.js # Contains console.log debugging
git commit -m "feat(app): add feature"
# ✓ GOOD: Review before committing
git diff --staged # Check for debug code
# Remove debug statements
git add src/app.js
git commit -m "feat(app): add feature"
# ✗ TOO LARGE: Entire feature in one commit
git commit -m "feat(auth): complete authentication system"
# 50 files changed, 2000+ lines
# ✗ TOO SMALL: Meaningless micro-commits
git commit -m "add semicolon"
git commit -m "fix typo"
git commit -m "add newline"
# ✓ GOOD: Right-sized atomic commits
git commit -m "feat(auth): add login endpoint"
git commit -m "feat(auth): add logout endpoint"
git commit -m "feat(auth): add token refresh mechanism"
git commit -m "test(auth): add authentication tests"
# ✗ DANGEROUS: Amending after push
git push origin feature-branch
# Oh, forgot a file!
git add forgotten.js
git commit --amend --no-edit
git push --force origin feature-branch # ← Causes problems!
# ✓ SAFE: New commit instead
git push origin feature-branch
# Oh, forgot a file!
git add forgotten.js
git commit -m "feat(auth): add missing validation helper"
git push origin feature-branch
feat(<scope>): <what you added>
<why this feature is needed>
<how it works (if not obvious)>
<any limitations or considerations>
Closes #<issue-number>
fix(<scope>): <what you fixed>
<what was wrong>
<how you fixed it>
<why this approach>
Fixes #<issue-number>
refactor(<scope>): <what you refactored>
No behavior change. <why refactor was needed>
<what improved (readability, performance, maintainability)>
When AI agents generate code that needs committing:
# ✓ CORRECT: Check what changed
git status
git diff
# Determine:
# - Is this one logical change or multiple?
# - What type of change? (feat, fix, refactor, etc.)
# - What scope? (auth, api, ui, etc.)
# ✓ GOOD: AI-generated message with context
git commit -m "feat(api): add rate limiting middleware
Implements token bucket algorithm with 100 requests/minute limit.
Returns 429 status with Retry-After header when limit exceeded.
Addresses security requirement from issue #789"
# ✗ BAD: Generic AI message
git commit -m "add rate limiting"
# If AI generated multiple files for different purposes:
# ✓ CORRECT: Separate commits
git add src/rate-limiter.js
git commit -m "feat(api): add rate limiting middleware"
git add tests/rate-limiter.test.js
git commit -m "test(api): add rate limiter tests"
git add docs/api.md
git commit -m "docs(api): document rate limiting"
# ✗ WRONG: One large commit
git add .
git commit -m "add rate limiting with tests and docs"
Before every commit, ask:
Good commit messages answer:
# ✓ EXCELLENT: Answers all questions
feat(search): implement fuzzy search algorithm
Users frequently make typos in search queries, resulting in zero
results and poor experience. Implemented Levenshtein distance-based
fuzzy matching with tolerance of 2 characters.
Performance impact: ~50ms additional latency for searches, acceptable
given improved user experience. Results are cached for 5 minutes.
Closes #456
Key Principles:
git add -p for partial staging)Quick Command Reference:
git add <file> # Stage specific file
git add -p <file> # Stage parts of file interactively
git diff --staged # Review what will be committed
git commit -m "type(scope): msg" # Commit with message
git commit --amend --no-edit # Amend last commit (if not pushed)
git log -1 --stat # Verify last commit
Remember: Commits are documentation of your project's history. Write them for humans who will read them 6 months from now.