claude-skills
CLAUDE.md1 starsChanged 19 days agoApache-2.0View on GitHub
A collection of custom Claude Code skills — reusable prompt packages, pure Markdown, no build/lint/test tooling. Each skill lives in its own top-level directory and is installed by copying that directory into ~/.claude/skills/ (global) or .claude/skills/ (per-project). This repo is a git submodule of the Metis monorepo (skills/claude-skills), with its own remote (seevali/claude-skills) and independent history. Commit and push here directly; the Metis parent repo tracks the submodule pointer separately.
In this repository3 other files it gives its agents.
- design-councilSkill
- mission-commandSkill
- ratification-docketSkill
The file
We split this CLAUDE.md into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
Every skill also gets a README.md written for humans (overview, installation commands, usage), and a row in the root README.md skills table.
- Very specific: line 33
When editing a skill's SKILL.md, keep its README.md and the root README.md table in sync — descriptions drift apart otherwise.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.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.
Files
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
What This Repository Is
A collection of custom Claude Code skills — reusable prompt packages, pure Markdown, no build/lint/test tooling. Each skill lives in its own top-level directory and is installed by copying that directory into ~/.claude/skills/ (global) or .claude/skills/ (per-project).
This repo is a git submodule of the Metis monorepo (skills/claude-skills), with its own remote (seevali/claude-skills) and independent history. Commit and push here directly; the Metis parent repo tracks the submodule pointer separately.
Skill Anatomy
<skill-name>/
├── SKILL.md # The skill itself — loaded by Claude when triggered
├── README.md # Human-facing docs: what it does, installation, usage
├── references/ # Optional: supporting material loaded on demand
└── .claude-plugin/
└── plugin.json # Plugin manifest (name, description, author, license, homepage, repository, version)- SKILL.md frontmatter is the trigger mechanism. The YAML
descriptionfield is what Claude reads to decide whether to activate the skill — it must enumerate trigger phrases, use cases, and scope explicitly (see either existing skill for the expected density). Thenamefield must match the directory name. - SKILL.md body is the payload. It's written as instructions to Claude, not documentation about the skill. Keep it lean; push long supporting material into
references/. - Every skill also gets a
README.mdwritten for humans (overview, installation commands, usage), and a row in the rootREADME.mdskills table. - This repo installs via two mechanisms, both dependent on the layout above:
npx skillsCLI (Vercel'sskillspackage): its standard search paths are the repo root andskills/-style folders, so it finds this repo's top-level<skill-name>/SKILL.mdfolders through the entries declared in.claude-plugin/marketplace.json(verified against the CLI README on 2026-10-16) — the marketplace file is load-bearing for both install paths. Do not restructure skills away from this shape.- Claude Code plugin marketplace:
.claude-plugin/marketplace.jsonat repo root lists each skill as a plugin entry (name,source,description,license); each skill dir carries its own.claude-plugin/plugin.jsonwith matching metadata plus a semverversion. The version lives inplugin.jsononly, never in the marketplace entry. A singleSKILL.mdat a plugin's root (noskills/subdirectory needed) is a documented default location — confirmed working viaclaude plugin validateon 2026-07-05.
Conventions
- License is Apache 2.0, copyright Seevali Rathnayake — new skills reference the root
LICENSE(see the "License" section pattern inmission-command/README.md). - Some tracked
*:Zone.Identifierfiles exist (WSL artifacts from copying files in from Windows). Do not create or commit new ones; ignore the existing ones unless asked to clean them up. - When editing a skill's SKILL.md, keep its README.md and the root README.md table in sync — descriptions drift apart otherwise.
- Bump a plugin's
versionin itsplugin.jsonwhenever you change anything in that skill's folder (patch for fixes and metadata, minor for new behaviour, major for breaking changes). Why: Claude Code compares versions to decide whetherclaude plugin update <name>@claude-skillsfetches anything. A version that never moves leaves installed users stuck on old code. That happened between 2026-10-15 and 2026-10-26, when 11 commits shipped under a fixed1.0.0. Droppingversionand letting the commit SHA stand in would also fix updates, but the HOL plugin scanner (hol.org, which gates the awesome-ai-plugins catalog listing) flags a missingversionas a medium finding, andclaude plugin validate --strictwarns about it. - A new skill needs four things kept in sync: its own
README.md, a row in the rootREADME.mdtable, a.claude-plugin/plugin.jsoninside the skill dir, and a matching entry in the root.claude-plugin/marketplace.jsonpluginsarray. Runclaude plugin validate .(andclaude plugin validate ./<skill-name>) after adding one.
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
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.

