agentleFS
zhangliang0115zhangliang0115/ai-plugin/skills/skill-author/SKILL.md

skill-author

Skill0 starsChanged 2 months agoMITView on GitHub

Write, structure, and publish a SKILL.md skill that works across every harness — Claude Code, DeepSeek Harness (dsh), Codex CLI, Gemini CLI, Copilot, Cursor — including the cross-agent gotchas generic guides miss: tier shadowing between project and user roots, discovery rules, and trigger-description design. Use when creating a new skill, fixing a skill that fails to load in some agents, or preparing a skill repo for distribution.

24 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

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 19 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Keep SKILL.md under ~500 lines: Put depth in references/*.md and link to them ("see references/schema.md").

    line 51
  2. Very specific:

    Relative links must stay relative: If SKILL.md references scripts/x.py, the whole folder travels together — never link outside the skill …

    line 56

“Common” is measured: kept by at least 6 in 10 of the 19 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

Authoring a cross-agent SKILL.md skill

A skill is a folder with a SKILL.md file inside. The file starts with a YAML frontmatter block (name + description) followed by Markdown instructions the agent loads on demand. Because Claude Code, DeepSeek Harness (dsh), Codex CLI, Gemini CLI, Copilot and Cursor all read this same format, one well-authored skill serves every agent.

Anatomy

my-skill/
├── SKILL.md            # required: frontmatter + instructions
├── references/         # optional: long-form docs the model reads on demand
└── scripts/            # optional: helper scripts the skill may run
---
name: my-skill
description: >-
  One or two sentences: what it does AND when to use it. This is the trigger
  the agent matches against the user's request — write it like a routing rule.
---

# My skill

Step-by-step instructions. Assume a competent engineer; be concrete, not vague.

Rules that decide whether a skill works

  1. name must be kebab-case — lowercase letters, digits, hyphens; max 64 chars. Agents reject or mangle other names. aipx normalizes for you, but ship it correct.
  2. description is the trigger, not documentation. Bad: "Utilities for working with things." Good: "Summarize YouTube/Bilibili videos and podcasts from a URL. Use when the user pastes a video or audio link."
  3. Keep SKILL.md under ~500 lines. Put depth in references/*.md and link to them ("see references/schema.md"). Agents load the body only when the skill triggers; huge bodies waste context and dilute behavior.
  4. No nested skills. Discovery scans only the direct children of a skill root — skills/a/SKILL.md is found, skills/a/b/SKILL.md is not.
  5. Relative links must stay relative. If SKILL.md references scripts/x.py, the whole folder travels together — never link outside the skill directory.

Where skills live (user roots)

RootRead by
~/.agents/skills/<name>/SKILL.mdDeepSeek Harness (dsh), Codex CLI
~/.claude/skills/<name>/SKILL.mdClaude Code
~/.gemini/skills/<name>/SKILL.mdGemini CLI
~/.copilot/skills/<name>/SKILL.mdGitHub Copilot CLI

Project-scoped equivalents exist too (e.g. .claude/skills/, .agents/skills/ in the repo). Don't copy by hand — use aipx sync to link one copy into every agent root.

Test before publishing

# does it parse and get discovered?
aipx install ./my-skill --dry-run
aipx install ./my-skill
aipx list

Then open your agent and confirm the skill appears (Claude Code: it shows in the skill list; dsh: / command palette → Skills group).

Publishing checklist

  • Repo contains the skill folder at a stable path, e.g. skills/my-skill/
  • description states trigger conditions, not just features
  • Works from a clean install (aipx install owner/repo#path:/skills/my-skill)
  • Add GitHub topics so discovery works: agent-skills, claude-code, dsh-plugin (if it also ships a dsh bundle), deepseek (if DS-related)
  • Optional: dual-target packaging — see the dsh-plugin-dev and claude-plugin-dev skills

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.