Skill creator
Authors new SKILL.md employees for this repo — interviews for the job-to-be-done, picks a slug, writes a trigger-rich description under 500 chars, fills the standard body sections, then tests triggering against five paraphrased asks and iterates. Use when the user says "make me a skill for X", "turn this workflow into a skill", "add a new employee to the team", or "my skill never triggers".From its SKILL.md
npx -y skills add alebgl77/claude-inc --skill skill-creatorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 8 stars8 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.9 KB, 874 tokens by cl100k_base, as published. Nobody here has run it
Skill Creator — Skill Smith
"Build your own skills"
When to use
- "Make me a skill for <recurring task>" — anything you've done three times deserves a SKILL.md
- "Turn this workflow into a skill" after a session that went well
- "Add a new employee to <department>" in this repo
- "My skill never triggers" / "the wrong skill fires" — description surgery
- Standardizing a process so the whole team runs it the same way
Workflow
- Interview for the job-to-be-done: the task, three real phrasings the user would type to ask for it, available inputs, the artifact that means "done", and what must never happen.
- Choose the slug: lowercase, hyphenated, specific (
api-changelog, nothelper); checkskills/for collisions. - Write the description before the body — it is the trigger. Third person, capability first, then "Use when ..." quoting the phrasings from step 1. Hard limit 500 characters.
- Write the body in this repo's format:
# Name — Role, quoted tagline,## When to use,## Workflow(5–9 executable steps),## Output format(code-block template),## Quality bar(checkboxes),## Example. - Save to
skills/<slug>/SKILL.md— directory name must equal frontmattername, exactly. - Trigger-test with five paraphrased asks: a synonym rewrite, a vague version, a jargon version, an adjacent-but-different task (must NOT fire), and the exact phrase. Judge each against the description alone.
- Iterate the description until at least 4/5 positives fire and the negative stays quiet — widen with user vocabulary, narrow with distinguishing nouns.
- Dry-run the workflow once end to end; rewrite any step that needed unstated context.
Frontmatter contract
---
name: <slug> # must equal the directory name, lowercase-hyphenated
description: <capability + "Use when ..." with real user phrases; third person; under 500 chars>
---
Body sections, in order: title with role, tagline quote, When to use, Workflow, Output format, Quality bar, Example.
Trigger-miss failure modes
- Describes implementation, not the job — "uses Playwright" instead of "browser-tests your app". Users ask for jobs.
- Zero quoted user phrases — the matcher has nothing to catch.
- Too generic — collides with sibling skills and the wrong employee shows up.
- Over 500 characters — the tail gets diluted exactly where your triggers live.
- First-person or imperative voice — reads like instructions, not a capability index.
Output format
skills/<slug>/SKILL.md (created)
Trigger test:
| Ask (paraphrase) | Should fire | Fired |
|---|---|---|
| <synonym rewrite> | yes | yes |
| <vague version> | yes | yes |
| <jargon version> | yes | yes |
| <adjacent task> | no | no |
| <exact phrase> | yes | yes |
Score: 5/5 — shipped
Quality bar
-
nameequals the directory slug, exactly - Description is third person, under 500 chars, and quotes real user phrasings
- Workflow steps are executable with zero unstated context
- Output format is a copy-paste template, not a description of one
- Trigger test scores at least 4/5 and the adjacent negative stays quiet
- Example shows one real ask and the artifact it produced
Example
Ask: "Turn our release-notes ritual into a skill."
Produced: skills/release-notes/SKILL.md — description quoting "write the release notes", "what changed this sprint", and "draft the changelog"; a 7-step workflow from git log to published notes; trigger test 5/5, with the adjacent ask "write API docs" correctly staying quiet.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 1 of the 12 instructions most skill authoring skills give in 874 tokens
Counted across 408 of the 422 authors here whose files we hold, read 2026-09-06
- Keep SKILL.md under 500 linesin 118 of 408, across 104 files
- Write skill instructions in imperative formin 44 of 408, across 37 files
- Move detailed reference material into references filesin 41 of 408, across 34 files
- Run init_skill.py to initialize new skillsin 40 of 408, across 33 files
- Draft assertions while test runs are in progressin 32 of 408, across 25 files
- Keep description under 1024 charactersin 31 of 408, across 18 files
- Run package_skill.py once the skill is completein 29 of 408, across 23 files
- Spawn with-skill and baseline runs in the same turnin 28 of 408, across 21 files
- Test added scripts by actually running themin 27 of 408, across 21 files
- Follow the creation steps in orderin 24 of 408, across 17 files
- Generate the eval viewer before evaluating outputs yourselfin 24 of 408, across 17 files
- Write the description in third personhere, and in 24 of 408, across 18 files
Said here and by no other author read
- Interview for the job-to-be-done
- Collect three real user phrasings
- Choose a specific lowercase-hyphenated slug
- Use the standard body sections in order
- Keep workflows to five through nine steps
- Trigger-test with five paraphrased asks
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.