agent-skills
CLAUDE.md0 starsChanged 6 days agoMITView on GitHub
Promoted from repeat review findings (each class seen in two separate reviews).
In this repository31 other files it gives its agents.
- ai-estimateSkill
- aso-craftSkill
- audit-reportSkill
- babysit-prsSkill
- blueprintSkill
- brag-docSkill
- business-briefSkill
- cagan-checkSkill
- code-craftSkill
- commit-reportSkill
- compassSkill
- cv-craftSkill
- dev-flowSkill
- grill-meSkill
- implementation-reviewSkill
- llm-wikiSkill
- parallel-worktreesSkill
- periscopeSkill
- portfolio-syncSkill
- pr-craftSkill
- problem-solvingSkill
- publish-skillSkill
- review-passSkill
- security-auditSkill
- semverSkill
- sentinelSkill
- seoSkill
- testing-philosophySkill
- to-prdSkill
- write-a-skillSkill
- wstg-security-testingSkill
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 6
Write each eval grader so a wrong output fails it: check content, not just that a heading or keyword exists, and probe every regex against …
- Very specific: line 7
When a skill reads or measures a diff, name the exact git command for the target it reviews (--staged, the merge-base range, untracked …
≈ “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
agent-skills: rules for changes
Promoted from repeat review findings (each class seen in two separate reviews).
- A review gate must not be satisfiable by the reviewed agent's own output (a stated reason, a plan edit it committed, commit order); route anything only the agent can vouch for to the user.
- Write each eval grader so a wrong output fails it: check content, not just that a heading or keyword exists, and probe every regex against a near-miss before committing.
- When a skill reads or measures a diff, name the exact git command for the target it reviews (
--staged, the merge-base range, untracked files), never a baregit diff. - State a policy only in the skill that owns it; other skills name that skill instead of describing its behavior.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.