agentleFS
yanzyuyuyanzyuyu/min-tok/skills/git-craft/SKILL.md

git-craft

Skill1 starsChanged 47 days agoApache-2.0View on GitHub

MANDATORY: Execute FIRST before running ANY git commands (init, add, commit, push). Pragmatic Git workflow, human-like commit authoring, and repository hygiene without AI slop. Do NOT skip this. Use to ensure zero credential/agent leaks, human-style commit logs, and clean git history.

  • Reads credentials
  • Commits and pushes

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. About 0 are common, 10 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    STRICT BAN: Never mention skill names, prompts, slash commands, or AI tools in commit messages (e.g., NEVER write *"using markdown-doc …

    line 26
  2. Very specific:

    Use clean, standard verbs: init, add, fix, update, clean, remove, move.

    line 44
  3. Very specific:

    Before making the very first commit in any project, ALWAYS create an airtight .gitignore:

    line 51
  4. Distinctive:

    Never blindly run git add .

    line 89
  5. Distinctive:

    Always check git status before committing.

    line 90
  6. Distinctive:

    If a file was committed by mistake, untrack it while keeping the local file:

    line 91
  7. Distinctive:

    Always verify that files committed on Windows don't introduce encoding artifacts into Unix/CI environments.

    line 102
  8. Distinctive:

    Before creating commits in any project:

    line 108
  9. Distinctive:

    Verify the active authenticated GitHub identity:

    line 109
  10. Very specific:

    Ensure repository user.name and user.email match that identity:

    line 113
  11. Distinctive:

    Before pushing the initial commit or publishing any public repository:

    line 124
  12. Very specific:

    Mandatory Hero Banner: Must include an aesthetic dark-mode vector SVG banner (banner.svg) at the repository root, centered at the top.

    line 127
  13. Very specific:

    Mandatory Visual Proof: Must include a dark-mode vector SVG terminal demo (terminal_demo.svg) for CLI/backend tools or a clean UI snapshot …

    line 129
  14. Very specific:

    Every public repository must have a root LICENSE file (MIT, Apache-2.0, etc.) with the correct copyright holder.

    line 133
  15. Distinctive:

    Never leave a GitHub repository homepage with an empty "Releases" sidebar:

    line 139
  16. Distinctive:

    Create an official release via GitHub CLI:

    line 145

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Pragmatic Git & Repository Hygiene (git-craft)

This skill guides the agent to manage Git repositories cleanly, reliably, and author commit messages that look 100% like they were written by a real, senior human developer—eliminating all traces of robotic "AI commit slop".


1. Human-Grade Commit Messages (Anti-AI-Slop)

The "AI Commit Smell" to Eliminate:

AI notoriously generates robotic, over-formalized commit messages:

  • BAD (AI Slop): feat: initial commit for crosscheck data reconciliation engine
  • BAD (AI Slop): docs: upgrade README with visual terminal snapshot and benchmark diffs using markdown-doc skill
  • BAD (AI Slop): chore: implement persistent rules via /learn slash command

The 4 Rules for Authentic Human Commits:

  1. Never Leak Agent Mechanics:
    • STRICT BAN: Never mention skill names, prompts, slash commands, or AI tools in commit messages (e.g., NEVER write "using markdown-doc skill", "created by antigravity", "via /learn"). Commits should look like a human engineer typed them in their terminal.
  2. Learn and Match the User's Typing Style:
    • Observe how the user speaks in the conversation (casual tech-bilingual, direct, punchy).
    • Reflect that tone naturally: short, clear, and grounded.
  3. Keep It Punchy (3 to 7 words):
    • Real developers don't write essays in the commit title:
      • Instead of: feat: initial commit for crosscheck data reconciliation engine $\rightarrow$ init crosscheck engine or first commit: crosscheck
      • Instead of: docs: upgrade README with visual terminal snapshot and benchmark diffs using markdown-doc skill $\rightarrow$ update readme with terminal demo and benchmark
      • Instead of: chore: untrack .agents skills from repo and update project tree in README $\rightarrow$ remove .agents from repo and clean readme tree
      • Instead of: fix: resolve UnicodeEncodeError charmap issue on Windows cp1252 legacy console $\rightarrow$ fix windows terminal encoding crash
  4. Natural Verb-First Phrasing:
    • Use clean, standard verbs: init, add, fix, update, clean, remove, move.
    • Optional lowercase prefixes (feat:, fix:, docs:) are fine ONLY IF kept brief and natural, not stuffed with prepositional fluff.

2. Bulletproof .gitignore from Day 1

Before making the very first commit in any project, ALWAYS create an airtight .gitignore:

# Python
__pycache__/
*.py[cod]
*$py.class
*.db
*.sqlite3
.pytest_cache/
venv/
.venv/

# Node / Web
node_modules/
dist/
build/
.next/

# Environment & Secrets
.env
.env.local
*.pem
*.key

# Agent / Tooling Configuration (NEVER push to product repos)
.agents/
.agent/
_agents/
_agent/

Critical Rule: Never commit internal agent workspaces (.agents/) into public or product repositories.


3. Atomic Staging & Untracking

  • Never blindly run git add . if sensitive files or untracked directories were accidentally created.
  • Always check git status before committing.
  • If a file was committed by mistake, untrack it while keeping the local file:
    git rm -r --cached <file_or_directory>

4. Windows Git & Encoding Hygiene

On Windows environments:

  • Be aware of CRLF vs LF warnings. Use git config core.autocrlf true (or standard repo .gitattributes).
  • Always verify that files committed on Windows don't introduce encoding artifacts into Unix/CI environments.

5. GitHub Account & Commit Author Verification

Before creating commits in any project:

  1. Verify the active authenticated GitHub identity:
    gh api user --jq '{login: .login, email: .email}'
  2. Ensure repository user.name and user.email match that identity:
    git config user.name "<github_login>"
    git config user.email "<github_email>"
    This ensures that all commits pushed to GitHub link directly to the user's verified profile and avatar, avoiding disconnected generic silhouettes.

6. Mandatory Repository Documentation Gate (README & Assets)

Before pushing the initial commit or publishing any public repository:

  1. Verify README.md Compliance:
    • The project must have a production-grade README.md based on markdown-doc/SKILL.md or templates in ~/.gemini/config/skills/markdown-doc/templates/.
    • Mandatory Hero Banner: Must include an aesthetic dark-mode vector SVG banner (banner.svg) at the repository root, centered at the top.
    • Mandatory Badges: Minimum 5 centered Shields.io badges immediately below the hero banner (Runtime, Release, Domain Metric, Dependencies, Test Suite/Audit, License).
    • Mandatory Visual Proof: Must include a dark-mode vector SVG terminal demo (terminal_demo.svg) for CLI/backend tools or a clean UI snapshot for web apps, saved directly at the repository root.
    • Mandatory Vector Back-to-Top: Navigation links at section ends must use the dark-mode vector SVG button (▲ Back to Top).
    • Zero AI Marketing Slop: Strict ban on emojis in headings and marketing hype.
  2. Verify License File:
    • Every public repository must have a root LICENSE file (MIT, Apache-2.0, etc.) with the correct copyright holder.

7. Semantic Versioning & GitHub Releases

Never leave a GitHub repository homepage with an empty "Releases" sidebar:

  1. Tag commits using Semantic Versioning (v1.0.0, v1.1.0, etc.):
    git tag -a v1.0.0 -m "v1.0.0: initial production release"
    git push origin v1.0.0
  2. Create an official release via GitHub CLI:
    gh release create v1.0.0 --title "v1.0.0: Initial Production Release" --notes-file RELEASE_NOTES.md
  3. Set the repository metadata:
    gh repo edit --description "<Precise technical summary>"
    gh repo edit --add-topic "<topic1>" --add-topic "<topic2>"

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.