agentleFS
VelimirMuellerVelimirMueller/ai-development-skills/skills/core/extend-skillset/SKILL.md

extend-skillset

Skill7 starsChanged 16 days agoMITView on GitHub

Use when the stack profile names a tool no installed skill covers, or the user asks for a new skill. Authors one in the house style, audit-first and profile-aware, into the project's .claude/skills or this plugin repo, then checks it.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are 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.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run the checklist at the end of skill-template.md.

    line 56

≈ “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.

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

Extend Skillset

Turns "no skill covers X" into a skill that fits the set. Template and checklist: skill-template.md. House style: this repo's CONTRIBUTING.md.

1. Audit (change nothing)

cat .claude/stack-profile.md 2>/dev/null
ls .claude/skills 2>/dev/null; ls skills/*/ 2>/dev/null | head -80
grep -rli "<tool or topic>" .claude/skills skills 2>/dev/null | head

Answer three questions before writing anything:

  1. Is there already a skill that covers it (even partly)? Extend that one; do not add a twin.
  2. Is the gap a procedure with a trigger ("when X, do Y")? If it is only a fact, add it to a reference file or _shared/ instead.
  3. Is the stack stable enough to encode? A tool in the "Assess" ring of tech-radar.md gets a short skill with a warning, not a full one.

2. Decide

SituationAction
Existing skill covers itEdit it; add the case; stop.
Fact onlyAdd a rule to the nearest reference file; stop.
New procedureContinue.

Choose the destination (ask once with AskUserQuestion, recommended first):

  • Project (Recommended): .claude/skills/<name>/ — private to this repo, zero release cost.
  • Upstream: skills/<catalogue>/<name>/ in this plugin repo — for skills useful to others. Needs registration by the maintainer (README, CHANGELOG, validator).

3. Research the stack live

Follow version-protocol.md. For each tool the skill touches: latest stable version, current install command, config keys, breaking changes of the last major. Read the official docs page, not a blog. Write each verified fact with its source into the reference file, and into the catalogue's _shared/stack-versions.md when one exists. If you cannot verify a fact, mark it unverified and say so in the skill.

4. Write the skill

Use skill-template.md. Requirements:

  • Folder <name> kebab-case, verb-first (set-up-x, configure-x, build-x, audit-x); name equals the folder.
  • description starts with Use when, one sentence, at most 300 characters, names the trigger situation, not the topic.
  • Body: numbered, audit-first, idempotent: Audit → Decide → Detect track → Install only the missing → Generate seams → Wire → Verify (exact command and expected output). At most about 180 lines.
  • Step 1 reads .claude/stack-profile.md and branches on it; if absent, it detects from the repo and asks one question only when detection is ambiguous and the answer changes the output (then suggests set-up-stack-profile).
  • Reference file: each rule as ## Rule: with Why, How to apply, optional Anti-example; ends with ## When to deviate.
  • Link to the shared contracts instead of restating them: engineering-principles.md, security-baseline.md, logging-contract.md, observability.md.
  • Code snippets compile against the current API. One canonical pattern, not a menu.
  • Plain prose; none of: simply, just, seamless, robust, leverage.

5. Check against the checklist

Run the checklist at the end of skill-template.md. In this plugin repo, also run:

bash scripts/validate.sh

Expected: OK: validator passed. Outside the repo, run the checklist's shell snippet, which reproduces the three checks (frontmatter, description rule, links).

6. Verify the skill triggers

Describe a realistic task in one sentence and check that the description would match it and not match an unrelated one. Rewrite the description if not. Then run the skill's own audit step on a real repo and confirm it changes nothing on a second run.

7. Report

Print: path written, the verified versions with sources, what is unverified, and (for upstream) the registration steps left for the maintainer: README list, CHANGELOG entry, manifest listing. Do not edit those yourself unless asked.

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.