agentleFS
Sign inStart free
jocarrdjocarrd/crux-agents/skills/skill-audit/SKILL.md

skill-audit

Skill4 starsChanged 18 days agoMITView on GitHub

Review whether any skill of a project is really a pattern that deserves moving up to the kernel, or whether any kernel skill has filled up with special cases and should move down. Inventories the skills of all projects, measures their density of proper names and detects pairs of different projects that solve the same thing. Use when creating a new skill, and periodically at session close.

58 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use write-a-skill, reflect and 7 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • simo97Skill

    write-a-skill

    Create new agent skills with proper structure, progressive disclosure, and bundled resources. Use when user wants to create, write, or build a new skill.

    Also in 4 projects that use this

  • Microsoft AzureSkill

    reflect

    Learning capture system that extracts HIGH/MED/LOW confidence patterns from conversations to prevent repeating mistakes. Use after user corrections ("no", "wrong"), praise ("perfect", "exactly"), or when discovering edge cases. Complements .squad/agents/{agent}/history.md and .squad/decisions.md.

    Also in 5 projects that use this

  • sunnyclinches44Skill

    explain

    Explains code using real-world analogies and ASCII diagrams

    Also in 4 projects that use this

  • eherrerosjSkill

    humanize

    Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive 'Signs of AI writing' guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases. Triggers on: 'humanize', 'make this sound human', 'remove AI writing', 'de-AI', 'sound more natural', 'less robotic', 'too AI-sounding', or when reviewing text that reads like AI output.

    Also in 3 projects that use this

  • MicrosoftSkill

    report

    Turn an exploration (threads, findings, charts) into a single Markdown report — note, blog post, executive summary, KPI dashboard, slide brief, or multi-section analytical report, with embedded charts.

    Also in 3 projects that use this

  • GoogleSkill

    grill-me

    A relentless interview to sharpen a plan or design.

    Also in 8 projects that use this

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 52 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When: when creating a skill (pattern or adapter?) and periodically at session close.

    line 14
  2. Distinctive:

    Write the kernel skill without projects or tools of any one agent.

    line 74
  3. Distinctive:

    Leave in the project an adapter with its own paths, credentials and gotchas that points to the kernel.

    line 75
  4. Distinctive:

    Run the new skill against the real case it came from: it has to give the same result.

    line 77

“Common” is measured: kept by at least 6 in 10 of the 52 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

skill-audit

It maintains the boundary between the kernel and the projects, and the short catalogue: procedures locked inside a project that should move up, kernel skills full of special cases that should move down, and skills that compete for the same sentence. The reasons and measurements behind each threshold: <this-skill-directory>/references/why.md.

When: when creating a skill (pattern or adapter?) and periodically at session close.

The test

Cover all the proper names in the skill. If the procedure still makes sense and has value, there is a pattern inside.

The procedure is promoted and the project keeps an adapter with paths, credentials and names. Two guards:

  • Two occurrences: it is promoted when two projects genuinely need it (except with demonstrated high value and almost no coupling).
  • Reverse path: a kernel skill with per-project conditionals moves down to project/.

Audit

node <this-skill-directory>/scripts/audit.mjs

It walks the adopted docs repos and reports:

  • Low density of proper names (derived from each project's crux.json): suspected of containing a pattern.
  • Pairs of different projects that describe the same thing: the two-occurrences rule is already met.
  • Those that are already adapters, separately and outside the candidates.
  • Adapters that compete with their procedure: descriptions that overlap by more than 0.18, or an adapter that does not name its project in the description.

They are signals, not verdicts: the script points, the test decides and the two-occurrences rule brakes. Size is watched by the description verifier: a skill of more than 1,000 words is shortened with /skill-creator.

Verify activation

node <this-skill-directory>/scripts/probe.mjs "the sentence you would ask for it with"
node <this-skill-directory>/scripts/probe.mjs --cases
node <this-skill-directory>/scripts/probe.mjs --adapter <adapter> <kernel> "<generic sentence>" "<own sentence>"
  • probe says which skill wins a sentence and by how much: below 15 % of the leader is a tie.
  • --cases runs casos/activacion.tsv (real sentence → expected skill). A sentence without proper names has to activate the kernel, never an adapter.
  • --adapter: the generic sentence is won by the kernel and the own one, by the adapter.

It is useful for comparing two wordings and for seeing whether a change switched a skill off. It does not predict the agent: a tie is reliable, a winner guarantees nothing. The sentence is written the way the job is asked for, not copied from the description.

Promote

  1. Separate the procedure from the data.
  2. Write the kernel skill without projects or tools of any one agent.
  3. Leave in the project an adapter with its own paths, credentials and gotchas that points to the kernel.
  4. Run the new skill against the real case it came from: it has to give the same result.
  5. Reinstall the kernel.

Anti-patterns

  • Promoting just in case: a bloated kernel is as bad as an empty one.
  • Promoting without leaving an adapter.
  • Ignoring the reverse path.

Related

  • /skill-creator — write or shorten a skill
  • /end-session — where the periodic review hooks in
  • /promote-to-docs — the equivalent for documentation

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free