make-skill
Skill2 starsChanged 3 months agoMITView on GitHub
Write a reusable recipe the assistant follows the same way every time. Use when a job repeats and only one person does it, when someone asks for a skill or a slash command, or when the decision tree lands on level 0.
8 other projects use a version of this.
Adopted by
Most recently taken up by
- diegoos/skillstheir own version, first seen Nov 8, 2026
- Sharqiewicz/skillstheir own version, first seen Nov 8, 2026
- sevenevesai/skillstheir own version, first seen Nov 1, 2026
- jared833/claude-code-hookstheir own version, first seen Oct 21, 2026
- dotnet/efcoretheir own version, first seen Oct 8, 2026
- sergeyklay/.agentstheir own version, first seen Sep 28, 2026
Other versions8 projects use a version of this.
Word for word · 1
- daveebbelaar/project-template★ 4100% same
Edited · 0
None found.
Wrote their own version · 7
- dotnet/efcore★ 15ksame name
- sergeyklay/.agents★ 6same name
- diegoos/skills★ 3same name
- jared833/claude-code-hooks★ 2same name
- Sharqiewicz/skills★ 1same name
- qiaoruntao/skills★ 0same name
- sevenevesai/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 34
Done when: every recorded answer is written into project/DECISIONS.md.
- Very specific: line 38
Create .agents/skills/<name>/SKILL.md and .agents/skills/<name>/agents/openai.yaml.
- Distinctive: line 40
Follow SKILL-RULES.md. It is short and it is what separates a skill that behaves the same every run from one that improvises.
- Very specific: line 56
Update project/SLICES.md with the skill as a built slice.
“Common” is measured: kept by at least 6 in 10 of the 7 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.
Files
Make a skill
Turn a job someone does by hand into a recipe their assistant follows.
A skill is a text file with instructions. It costs nothing, works in five minutes, and needs no website, no database, and no monthly bill. For a job that repeats and that only they do, it beats software.
1. Watch them do the job
Have them walk through the last real time they did it. Not the general shape, the actual last time.
Ask what they opened, what they looked at, what they decided, and what came out at the end. Where they say "and then I check whether it's worth chasing", stop and find out how they decide that. The judgement is the part worth writing down; the clicking is not.
Done when: you can describe the job start to finish, including every decision point, and they agree that description is right.
2. Find the leading word
One word this job is about, that they already use. Chase, reconcile, triage, draft, quote.
The right word carries the meaning by itself, so the skill is shorter and the assistant reaches for it more reliably. It goes in the name, in the description, and through the body.
Done when: one word is chosen and it is a word they used unprompted.
3. Ask what varies
Two or three questions, no more, about what changes between runs. What comes in, what goes out, what makes a run unusual.
The answer decides how much the skill has to handle. A job that is the same every time is a short skill. A job with three genuinely different shapes needs those three named.
Done when: every recorded answer is written into project/DECISIONS.md.
4. Write the skill
Create .agents/skills/<name>/SKILL.md and .agents/skills/<name>/agents/openai.yaml.
Follow SKILL-RULES.md. It is short and it is what separates a skill that behaves the same every run from one that improvises.
Done when: both files exist, the skill has numbered steps, and every step ends on a condition you could check.
5. Run it
Have them run it, in their own assistant, on a real instance of the job. Not a made-up example.
Watch what it does. Where it guesses, the skill was vague there; sharpen that step and run it again.
Done when: they have run it on real work and the output is something they would actually use.
6. Write it down
Append to project/DECISIONS.md: what the skill does, why a skill rather than a website, and what would change that.
Update project/SLICES.md with the skill as a built slice.
Done when: both files are updated and they know the one thing that would move them to level 1: the day a second person needs to use this.
Where it lives
.agents/skills/<name>/ in this repository. Codex reads that path directly, and Claude Code reaches the same folder through the .claude/skills symlink.
To use it in every project rather than this one, copy the folder to ~/.agents/skills/ for Codex and ~/.claude/skills/ for Claude 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.