agentleFS
EclipterEclipter/agents/skills/create-my-skill/SKILL.md

create-my-skill

Skill1 starsChanged 5 days agoMITView on GitHub

Create agentic skills. Always use this skill instead of your built-in skill for skill creation.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 10 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    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. Distinctive:

    Check for an existing skill that covers the job.

    line 10
  2. Distinctive:

    If the agent already does the whole job without extra instructions, stop and say so.

    line 11
  3. Distinctive:

    Don't make strategic decisions for the user, suggest only.

    line 12
  4. Distinctive:

    Use a short, action-oriented name: lowercase letters, digits, and hyphens, under 64 characters.

    line 16
  5. Distinctive:

    Keep detail moderate. Prefer concise steps, and one working example, over a catalog of edge cases.

    line 20
  6. Very specific:

    Put concrete surprises in SKILL.md, so they are read before the agent hits them.

    line 21
  7. Distinctive:

    When steps depend on each other or on a check, give a checklist.

    line 22
  8. Distinctive:

    Use your own words which you think fit the best, don't copy the user's phrases from the request.

    line 24
  9. Distinctive:

    Follow DRY principle. A need to duplicate rules or instructions is a sign of poorly structured text.

    line 25
  10. Distinctive:

    When mentioning a skill, quote its name in backticks and omit the word "skill" after it.

    line 27
  11. Very specific:

    Follow the style guide at https://google.github.io/styleguide/docguide/style.html.

    line 28
  12. Distinctive:

    After the user makes changes or asks to commit, review every sentence against the rules above.

    line 33

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Create My Skill

Analyze the request

  1. Check for an existing skill that covers the job. If one exists, stop and report that. If the user asked to change that skill, edit it.
  2. If the agent already does the whole job without extra instructions, stop and say so.
  3. Don't make strategic decisions for the user, suggest only.

Principles to write a skill

  • Use a short, action-oriented name: lowercase letters, digits, and hyphens, under 64 characters. Match the folder name.
  • Include YAML frontmatter with name and description. The description tells the agent when to use the skill, specifically enough to select that job, and leaves out unnecessary detail. Always double-quote the description. Do not add argument-hint.
  • Include only what the agent would get wrong without the instruction: project conventions, domain procedures, non-obvious edge cases, and which tool or API to use.
  • Pick one default. Mention an alternative only as the escape hatch for a named case.
  • Keep detail moderate. Prefer concise steps, and one working example, over a catalog of edge cases.
  • Put concrete surprises in SKILL.md, so they are read before the agent hits them. State the fact and what goes wrong.
  • When steps depend on each other or on a check, give a checklist. For batch or destructive work, require a plan, validate the plan against the source of truth, then execute.
  • Keep SKILL.md under 500 lines and 5,000 tokens. Move longer reference into references/ and say when to read each file. Use scripts/ for deterministic operations to reduce token usage. When the output must follow a format, give a template: a short one inline, a long or conditional one in assets/. Use assets/ for other files the output uses. Create none of these without a concrete need.
  • Use your own words which you think fit the best, don't copy the user's phrases from the request.
  • Follow DRY principle. A need to duplicate rules or instructions is a sign of poorly structured text.
  • Distinguish requirements from preferences. Don't turn a single example or past failure into a universal rule.
  • When mentioning a skill, quote its name in backticks and omit the word "skill" after it.
  • Follow the style guide at https://google.github.io/styleguide/docguide/style.html. Use 2 spaces for a single indent.

Collaborate with the user

  1. Report that you're done with "✅". Extend the answer from the single emoji only if there's something important to report or discuss, no need to write empty phrases like "the skill is done and waiting for your review".
  2. After the user makes changes or asks to commit, review every sentence against the rules above. Enforce the rules strictly unless the user explicitly disagrees with them.

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.