agent-skill-audit
Skill0 starsChanged 13 days agoMITView on GitHub
Validate a skill library for metadata correctness, discoverability, duplicate naming, risky instructions, and portability. Use when adding, reviewing, packaging, or publishing agent skills.
In this repository25 other files it gives its agents.
- AGENTS.mdAGENTS.md
- api-contract-checkSkill
- build-rescueSkill
- change-narrativeSkill
- context-handoffSkill
- data-quality-gateSkill
- dataset-workshopSkill
- dependency-careSkill
- docs-syncSkill
- experiment-ledgerSkill
- failure-forensicsSkill
- model-diff-reviewSkill
- model-eval-designSkill
- multi-agent-crosscheckSkill
- parallel-task-designSkill
- performance-proofSkill
- release-readinessSkill
- repo-scoutSkill
- research-evidence-mapSkill
- review-lensSkill
- safe-migrationSkill
- security-surfaceSkill
- spec-firstSkill
- sql-integrity-checkSkill
- test-driven-changeSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 7 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
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
- Very specific: line 11
Check that name uses lowercase letters, digits, and hyphens, matches the directory name, and stays within the standard limit.
- Distinctive: line 12
Look for overlapping triggers, duplicated skills, missing prerequisites, nonportable absolute paths, undocumented scripts, and …
- Distinctive: line 14
Check that workflows tell the agent to distinguish observed facts from assumptions and to report commands/tests actually performed.
- Distinctive: line 15
Ensure dependencies and optional resources are documented.
- Distinctive: line 16
Run repository validation and inspect a sample package listing before release.
- Distinctive: line 22
Do not modify a skill to loosen safety restrictions merely to improve its apparent capability.
- Distinctive: line 23
Do not claim compatibility has been tested on a host that was not available.
- Distinctive: line 24
Treat natural-language instructions as executable policy: inspect them with the same seriousness as code.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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 Skill Audit
Procedure
- Find each skill folder and confirm it contains one
SKILL.mdwith YAML frontmatter and a focused body. - Check that
nameuses lowercase letters, digits, and hyphens, matches the directory name, and stays within the standard limit. Check thatdescriptionexplains the workflow and trigger conditions without being vague. - Look for overlapping triggers, duplicated skills, missing prerequisites, nonportable absolute paths, undocumented scripts, and agent-specific assumptions hidden in generic instructions.
- Audit instructions and scripts for secret access, data deletion, network calls, external uploads, privilege escalation, destructive commands, and implicit merge/deploy/publish permissions.
- Check that workflows tell the agent to distinguish observed facts from assumptions and to report commands/tests actually performed.
- Ensure dependencies and optional resources are documented. Package each skill independently when the target host expects a skill ZIP.
- Run repository validation and inspect a sample package listing before release.
Deliverable
A prioritized findings list, invalid metadata list, risk assessment, portability notes, and package/validation results.
Boundaries
- Do not modify a skill to loosen safety restrictions merely to improve its apparent capability.
- Do not claim compatibility has been tested on a host that was not available.
- Treat natural-language instructions as executable policy: inspect them with the same seriousness as code.
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.

