skills
CLAUDE.md2 starsChanged 5 days agoMITView on GitHub
Skills are organized into bucket folders under skills/: Every skill in engineering/, meta/, agent-building/, or video-editing/ (the promoted buckets) must be listed in its bucket README.md, in the top-level README.md, and in .claude-plugin/plugin.json's skills array. Skills in in-progress/ and deprecated/ appear only in their bucket README.md. Each skill entry links the skill name to its SKILL.md. Bucket and top-level READMEs group entries into User-invoked and Model-invoked. Every SKILL.md is either user-invoked (disable-model-invocation: true plus policy.allow_implicit_invocation: false in agents/openai.yaml, reachable only…
In this repository10 other files it gives its agents.
- dirt-builderSkill
- environment-designSkill
- video-publish-workflowSkill
- add-skillSkill
- create-reel-editing-briefSkill
- do-a-fair-cutSkill
- do-a-rough-cutSkill
- grill-me-reel-editSkill
- instagram-reel-editorSkill
- reel-asset-boardSkill
The file
We split this CLAUDE.md into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 10
Every skill in engineering/, meta/, agent-building/, or video-editing/ (the promoted buckets) must be listed in its bucket README.md, in …
- Very specific: line 18
Run scripts/validate.sh before every commit.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 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
Skills are organized into bucket folders under skills/:
engineering/: code workmeta/: skills that maintain this repoagent-building/: building agents and skillsvideo-editing/: the reel editing pipelinein-progress/: beta, public on purpose, not shipped in the plugindeprecated/: no longer used, not shipped, not linked
Every skill in engineering/, meta/, agent-building/, or video-editing/ (the promoted buckets) must be listed in its bucket README.md, in the top-level README.md, and in .claude-plugin/plugin.json's skills array. Skills in in-progress/ and deprecated/ appear only in their bucket README.md.
Each skill entry links the skill name to its SKILL.md. Bucket and top-level READMEs group entries into User-invoked and Model-invoked.
Every SKILL.md is either user-invoked (disable-model-invocation: true plus policy.allow_implicit_invocation: false in agents/openai.yaml, reachable only by the human) or model-invoked (omit both). The test for model-invoked: could the model usefully reach for this autonomously? A user-invoked skill description is a short human-facing summary; a model-invoked description carries "Use when..." triggers. Every skill has an agents/openai.yaml with interface.display_name and interface.short_description. Dependencies between skills are an explicit instruction to call the Skill tool with the name, and only model-invoked skills can be called that way.
No em-dashes in this repo's prose (SKILL.md, READMEs, scripts, JSON, YAML, comments). Rewrite the sentence with a comma, colon, period, parentheses, or a conjunction; never substitute the character blindly. The one exception: skills/agent-building/ and skills/video-editing/ are imported verbatim from other repos and keep their original prose, so scripts/validate.sh skips them.
Run scripts/validate.sh before every commit. It checks frontmatter, agents/openai.yaml, README and plugin registration, and the em-dash rule. Run claude plugin validate . --strict after touching either file in .claude-plugin/.
To add a skill, use /add-skill. To link every skill outside deprecated/ into ~/.claude/skills and ~/.agents/skills, run scripts/link-skills.sh (--dry-run previews). Entries are symlinks, so git pull keeps them current.
Human-facing PDFs live under docs/<bucket>/.
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.

