claude-skills
CLAUDE.md0 starsChanged 23 days agoMITView on GitHub
Claude Code skills, each in skills/<skill-name>/, released on its own.
- Commits and pushes
In this repositoryOne other file it gives its agents.
- create-planSkill
The file
We split this CLAUDE.md into the rules it gives an agent
4 rules. About 0 are common, 0 are distinctive, and 4 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.
4
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
- Very specific: line 13
Hooks: never bypass them (--no-verify).
- Very specific: line 24
The gate: never open or merge a PR for a branch that changes skills/<skill-name>/ until that skill has been evaluated.
- Very specific: line 25
Who starts it: when the user says the branch is ready to merge, ask whether to run evaluate my <skill-name> skill with skill-creator.
- Very specific: line 33
Never create tags or split/* branches by hand.
≈ “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-skills
Claude Code skills, each in skills/<skill-name>/, released on its own.
Git
- Base branch:
main, the only long-lived branch. - Branch names:
<type>/<skill-name>-<desc>for one skill,<type>/<desc>for repository work.<type>isfeat,bugfix,hotfixorchore. - Commits: Conventional Commits,
type(scope): summary. The scope is a skill folder name or one ofrepo,ci,deps,releaseandmain. Thecommit-msghook checks every message. - Skill branches (
<type>/<skill-name>-<desc>, and any branch that changesskills/) reachmainonly through a pull request. Push the branch, open the PR withgh pr create --base main, then stop. The user merges it on GitHub with "Create a merge commit". - Repository branches (
<type>/<desc>): oncepnpm testpasses, merge withgit switch main && git merge --no-ff --no-edit <branch>and push withgit push origin main. - Never force-push
main. - Hooks: never bypass them (
--no-verify).
Commands
- Install:
pnpm install - Tests:
pnpm test - Benchmark check:
pnpm run check:benchmarks. Thepre-pushhook and CI run it. - Record an evaluation:
pnpm eval:record <skill-name> <iteration-dir>
Skill evaluations
- The gate: never open or merge a PR for a branch that changes
skills/<skill-name>/until that skill has been evaluated. - Who starts it: when the user says the branch is ready to merge, ask whether to run
evaluate my <skill-name> skill with skill-creator. Never start an evaluation on your own. - Order: first write or update the test cases in
skills/<skill-name>/evals/evals.jsonwith the user, then run the loop, then record the result withpnpm eval:recordand commit it. - Never commit
skills/*-workspace/.
Releases
- release-please opens one release PR per skill, and the user merges it.
- A merged release creates the tag
<skill-name>-vX.Y.Zand the split tagsplit/<skill-name>/vX.Y.Z. - Never create tags or
split/*branches by hand.
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.

