agentleFS
zhaozhiwenzhaozhiwen/agent_skills/AGENTS.md

agent_skills

AGENTS.md0 starsChanged 19 days agoMITView on GitHub

Public repo (github.com/zhaozhiwen/agent_skills) collecting agent skills authored or adapted by Zhiwen Zhao, in the open Agent Skills format — usable by any compliant agent (Claude Code, Codex, Gemini CLI, Copilot, Cursor, OpenCode, ...), not tied to one client. This file is the canonical rules file; CLAUDE.md only imports it (@AGENTS.md) for Claude Code. This repo is the source of truth. The default install is bash install.sh, which symlinks each skill into both ~/.agents/skills/ (cross-client convention) and ~/.claude/skills/ (Claude Code). Never…

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository2 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Create or edit skills/<name>/ (edit in place — the symlinks make it live immediately).

    line 52
  2. Distinctive:

    Add/update its row in README.md.

    line 54
  3. Very specific:

    Run bash install.sh for new skills (creates both links.

    line 55
  4. Very specific:

    Commit and push only with explicit approval, each time.

    line 68

≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md 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
AGENTS.md

agent_skills — rules

Public repo (github.com/zhaozhiwen/agent_skills) collecting agent skills authored or adapted by Zhiwen Zhao, in the open Agent Skills format — usable by any compliant agent (Claude Code, Codex, Gemini CLI, Copilot, Cursor, OpenCode, ...), not tied to one client.

This file is the canonical rules file; CLAUDE.md only imports it (@AGENTS.md) for Claude Code.

This repo is the source of truth. The default install is bash install.sh, which symlinks each skill into both ~/.agents/skills/ (cross-client convention) and ~/.claude/skills/ (Claude Code). Never keep real copies in those dirs — edit here; the links make it live.

Layout

skills/<name>/SKILL.md        required; the `skills/*/SKILL.md` path is what installers discover
skills/<name>/scripts/        optional: executable code
skills/<name>/references/     optional: docs loaded on demand
skills/<name>/assets/         optional: templates, data, images
skills/<name>/evals/          optional: evals.json + fixtures (agentskills.io eval convention)
scripts/validate.py           checks every skill against the spec
install.sh                    default install: symlinks every skills/*/ into ~/.agents/skills + ~/.claude/skills
README.md                     one row per skill
LICENSE                       MIT

Nothing else at the top level without a reason recorded here.

Skill rules (Agent Skills spec)

  • name: 1–64 chars, [a-z0-9-], no leading/trailing/double hyphen, equal to the dir name.
  • description: ≤1024 chars; say what it does and when to use it, with trigger keywords.
  • Frontmatter: only spec fields — name, description, license, compatibility (≤500), metadata (string→string), allowed-tools (experimental). Every skill sets license: MIT and metadata.author, because installers copy only the skill folder and the repo LICENSE does not travel with it.
  • Third-party skills (adapted, translated, or vendored): metadata.author is the original author, metadata.upstream is a URL pinned to the source commit, our role goes in metadata.translator/metadata.maintainer, and skills/<name>/LICENSE keeps the upstream copyright notice verbatim. Record the origin in README.md's Credits section. Only MIT-compatible upstreams.
  • Client-specific extras (e.g. Claude disable-model-invocation, Codex agents/openai.yaml) only if the skill still works on clients that ignore them.
  • SKILL.md body < 500 lines / ~5000 tokens; push detail into references/, one level deep.
  • Relative paths from the skill root only; no absolute paths, no assumption of a specific agent's tools (say "read the file", not a tool name).
  • Scripts: non-interactive, --help, errors to stderr, declare dependencies (PEP 723 for Python) or state them in compatibility.

Adding / changing a skill

  1. Create or edit skills/<name>/ (edit in place — the symlinks make it live immediately).
  2. python3 scripts/validate.py must pass.
  3. Add/update its row in README.md.
  4. Run bash install.sh for new skills (creates both links; existing links are refreshed).

Never commit

  • Secrets, tokens, passwords, real emails, or absolute paths to private data.
  • ~/.claude/skills/synced/* — those are Anthropic-synced skills, not ours.
  • Caches (__pycache__, .DS_Store) and eval run outputs (<name>-workspace/).

Anything committed here is public. Scan new files before committing: grep -rniE 'token|secret|password|api[_-]?key|/home/' skills/

Git

Commit and push only with explicit approval, each time. Tag releases (vX.Y.Z): gh skill install uses the latest tag.

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.