agentleFS
Sign inStart free
educlopezeduclopez/mallard/skills/skill-author/SKILL.md

skill-author

Skill4 starsChanged 13 days agoMITView on GitHub

Guided way to create or edit a skill, command or agent in the mallard repo: frontmatter, trigger phrases in Spanish, a "what it does not do" section, version bump rule, writing pass, lint. Use when the user says "crea una skill", "añade una skill a mallard", "mejora la descripción de la skill", "esta skill no se dispara", "nueva skill para X", or edits anything under skills/ or claude/.

23 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 3

    Distinctive

    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
  • could be automatic: a setting could enforce it
  1. Distinctive:

    If a skill with the same job exists, edit it instead of adding another.

    line 25
  2. Very specific:

    Add the skill to the README table, then run scripts/lint-skills.sh.

    line 61
  3. Distinctive:

    Test the trigger: read the description as the model would and check that three real phrasings would fire it and one unrelated phrasing …

    line 62
  4. Distinctive:

    Never commit, tag or release without the user asking.

    line 64

“Common” is measured: kept by at least 6 in 10 of the 18 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-author

Adapted from pstack's skill-creation flow plus technical-writing and unslop. The release steps live in mallard-release and docs/adding-skills.md; this skill covers the writing.

1. Decide the shape

  • Repeatable multi-step workflow with judgment: skill in skills/<name>/SKILL.md.
  • Short entry point the user types (/name): command in claude/commands/<name>.md.
  • Isolated expert with its own tools: agent in claude/agents/<name>.md.
  • A rule that must always hold: not a skill. Encode it as a lint (scripts/lint-skills.sh) or a CI check, and list it in docs/skill-rules.md.

If a skill with the same job exists, edit it instead of adding another.

2. Frontmatter

---
name: <dir-name>
description: >
  What it does, then "Use when the user says "..."" with 4-6 real phrases in
  Spanish (and English if the user mixes), plus the contexts and file names that
  should fire it.
version: "0.1.0"
metadata:
  author: <name>
---

The description is the only text that decides whether the skill loads. Put triggers there, not in the body.

3. Body

  1. One sentence: what it does and the closest skill it is not (does not set X up, use Y).
  2. Numbered steps with real commands. Imperative voice, one action per step, one term per thing.
  3. A table only when comparing outcomes (symptom, cause, action).
  4. A "what it does not do" line. Include any action that needs the user's OK.
  5. Long reference material goes to references/*.md, linked from the body.

4. Writing pass

  • Cut filler: "it is important to", "make sure to", opening summaries, closing recaps.
  • Short sentences, active voice, no invented jargon. Keep commands and paths exact.
  • Spanish tics to drop: "cabe destacar", "es importante señalar", "en definitiva", stacked adverbs in "-mente".

5. Version and lint

  • New skill: 0.1.0. Wording or small fix: patch. New step or flow: minor. Behavior change that breaks users: major.
  • Add the skill to the README table, then run scripts/lint-skills.sh. Fix every error before reporting done.
  • Test the trigger: read the description as the model would and check that three real phrasings would fire it and one unrelated phrasing would not.

Never commit, tag or release without the user asking.

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