Skip to content
Back to skills

Commit Messages

ASecurity

Use this skill to draft a commit message for staged changes. Execute `Skill(sanctum:git-workspace-review)` first to capture the repository path, status, and diffs. If no changes are staged, stage the relevant files before continuing.

  • 666 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added February 8, 2026
documentationgobashrailsgit

Works with

  • cli

Security analysis

A100/100

Scanned February 12, 2026

npx -y skills add majiayu000/claude-skill-registry --skill commit-messages --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Commit Messages?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Commit Messages
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majiayu000-commit-messages/badge)](https://www.skillsdirectory.com/skills/majiayu000-commit-messages)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: commit-messages
description: |

Triggers: messages, conventional-commits, git, changelog, commit
  Generate conventional commit messages from staged changes with correct type/scope.

  Triggers: commit message, conventional commit, git commit
  Use when: generating commit messages in conventional commits format
  DO NOT use when: full PR preparation - use pr-prep instead.
category: artifact-generation
tags: [git, commit, conventional-commits, changelog]
tools: [Bash, Write, TodoWrite]
complexity: low
estimated_tokens: 600
dependencies:
  - sanctum:shared
  - sanctum:git-workspace-review
version: 1.3.7
---

# Conventional Commit Workflow

## Usage

Use this skill to draft a commit message for staged changes. Execute `Skill(sanctum:git-workspace-review)` first to capture the repository path, status, and diffs. If no changes are staged, stage the relevant files before continuing.

## Required Steps

1. **Validate Code Quality**: Run `make format && make lint`. Fix any reported errors before proceeding. Do not bypass pre-commit hooks with `--no-verify` or `-n`.
2. **Classify the Change**: Choose a type from `feat`, `fix`, `docs`, `refactor`, `test`, `chore`, `style`, `perf`, or `ci`. Select an optional but preferred scope (e.g., `core`, `cli`). Identify breaking changes for the `BREAKING CHANGE:` footer.
3. **Draft the Message**:
    - **Subject**: `<type>(<scope>): <imperative summary>` (≤50 characters).
    - **Body**: Wrap at 72 characters. Explain the "what" and "why" behind the change. Use paragraphs for technical rationale.
    - **Footer**: Include breaking change details or issue references.
4. **Write the Output**: Use a relative path (e.g., `./commit_msg.txt`) to save the message. Overwrite the file with the final message only, without commentary.
5. **Preview**: Display the file contents using `cat` or `sed` for confirmation.

## Guardrails

Do not use `git commit --no-verify` or the `-n` flag. Pre-commit hooks are mandatory; fix issues rather than bypassing them. Commit messages must not include AI attribution or filler words like "leverage" or "seamless." Use a plain-text, human voice with present-tense imperative style for the subject line. Multi-line bodies are required for complex changes to explain the technical reasoning.

## Technical Integration

Combine this skill with `Skill(imbue:catchup)` or `/git-catchup` if additional context is needed. If the type or scope is unclear, re-examine the diffs or consult the project plan before finalizing the draft.

## Troubleshooting

Address specific errors reported by pre-commit hooks. Run `make format` to resolve styling issues and `make lint` for logic or style violations. Fix all detected issues before re-attempting the commit. If a merge conflict occurs, use `git merge --abort` to return to a clean state. A rejected commit indicates a failed quality gate; analyze the output and apply fixes before retrying.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.