dictionary
Skill14 starsChanged 2 months agoMITView on GitHub
A named, reusable procedure the agent loads on demand: a folder of instructions and optional scripts, used when a task matches. §06 Memory & Steering · also called: Agent skill, SKILL.md, Skills · interactive version A skill is a packaged procedure an agent can pick up when a task calls for it. In practice it is a folder with a SKILL.md describing what the skill does and when to use it, followed by the step-by-step instructions, and sometimes scripts or…
2 other projects use a version of this.
Adopted by
Most recently taken up by
- mattpocock/dictionary-of-ai-codingtheir own version, first seen Oct 8, 2026
- jimyoshida/generic-agent-skillstheir own version, first seen Jul 22, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- mattpocock/dictionary-of-ai-coding★ 5ksame name
- jimyoshida/generic-agent-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
Files
<!-- GENERATED FILE — do not edit by hand.
Source: lib/courses.ts in the Vibe Code School site (scripts/export-oss.mts).
Content PRs are welcome: edits get applied to the site source and re-exported. -->Skill
A named, reusable procedure the agent loads on demand: a folder of instructions and optional scripts, used when a task matches.
§06 Memory & Steering · also called: Agent skill, SKILL.md, Skills · interactive version
A skill is a packaged procedure an agent can pick up when a task calls for it. In practice it is a folder with a SKILL.md describing what the skill does and when to use it, followed by the step-by-step instructions, and sometimes scripts or templates alongside. The harness shows the agent only the name and description until the skill is invoked, then loads the body into context. That makes skills a working example of progressive disclosure.
Skills exist because repeated instructions rot in chat. A deploy procedure you explain in a prompt is gone when the session ends, and the version you paste next week drifts from the version you pasted last week. Putting it in a skill gives it a home, a name, and a single place to fix. The failure mode is the same as with AGENTS.md: skills that try to do too much, or ten near-duplicate skills whose descriptions blur together so the agent picks the wrong one.
Write a skill for anything you have explained twice. Keep the description specific enough that the agent can tell when it applies. Put deterministic steps in scripts the skill can run, and keep the prose for judgment calls. Skills differ from a slash command, which you trigger by hand, in that the agent can choose to use a skill itself, and from a subagent in that a skill runs inside the current session's context rather than in a separate one.
In the tools
- Claude Code: skills live in
.claude/skills/<name>/SKILL.md(project) or~/.claude/skills/(personal), invoked by name or chosen by the agent. - Claude Cowork: skills teach it your way of doing a job, such as how your team formats a weekly report.
- Codex: recent versions read the same
SKILL.mdconvention from a skills directory; check your version's docs for the path.
In conversation
“Every release I re-explain the changelog format and it still gets it slightly wrong.”
“Make it a skill. Description: 'writing release notes'. Body: the format, the sections, one example.”
“Do I have to invoke it?”
“You can with a slash, but it'll also pick it up on its own when you say 'draft the release notes'.”
Related: MCP · Hooks · Context engineering
Learn it in the school
Part of the AI Coding Dictionary by Vibe Code School · read this entry on the site · added 2026-09-12
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.

