agentleFS
jrgilbertsonjrgilbertson/the-rookery/skills/creating-portable-skills/SKILL.md

creating-portable-skills

Skill15 starsChanged 13 days agoMITView on GitHub

Create, revise, migrate, or audit portable Agent Skills packages. Use when the user requests skill authoring or explicitly invokes this skill to review instructions, structure, or portability.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep one instruction set for the job.

    line 31
  2. Very specific:

    Use skills-ref validate <skill-directory> from an available trusted install.

    line 43
  3. Distinctive:

    If it is unavailable, check the specification's frontmatter rules, name-directory match, body, and local resource links directly, and …

    line 44
  4. Distinctive:

    Check the final package's local links, dependencies, host conventions, and applicable evaluation evidence.

    line 55

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Creating Portable Skills

Produce an installable package that does the user's reusable job and follows the Agent Skills specification. Add behavioral evals when observed failures warrant them under the host's evaluation policy.

For a read-only audit, apply references/review-checklist.md and return prioritized findings with file evidence and a verdict. For authorized creation, revision, or migration, use the workflow below. Existing authorization and the host project's constraints carry forward.

Author the package

Ground the skill in the user's intent, existing package, representative inputs and outputs, and corrections from real work. Identify the outcome and hard constraints that make completion acceptable. Check the host's contribution rules, discovery path, and existing owners before adding a workflow or helper.

For a new skill, use assets/skill-template.md. For a revision, the last commit can preserve the prior version. For a migration, revise a copy and preserve the source. Read references/portability.md when choosing frontmatter, bundling dependencies, or removing harness-specific assumptions.

Keep one instruction set for the job. Separate changing facts from reusable instructions when that lets the next update change one source. Put conditional guidance in references/, copied outputs in assets/, and deterministic helpers in scripts/ when an existing mechanism cannot meet the requirement. Bundle everything the package needs, or declare its actual environment dependencies.

Apply references/review-checklist.md while drafting. Preserve user authority, exact formats, privacy requirements, and fragile operation order; leave reasoning and implementation choices open.

Validate and evaluate

Use skills-ref validate <skill-directory> from an available trusted install. If it is unavailable, check the specification's frontmatter rules, name-directory match, body, and local resource links directly, and record the limitation. Exercise bundled helpers with realistic inputs in disposable storage outside the host project; preserve existing inputs and outputs.

Package checks, helper tests, and applicable authorized smoke checks run for every package. For behavioral evaluation, follow references/evaluation.md.

Deliver

Check the final package's local links, dependencies, host conventions, and applicable evaluation evidence. Report the package location, observed checks, and anything unverified. For public migrations, replace private examples with synthetic inputs that preserve the behavior and difficulty being tested, and inspect the whole copy for leaks. Keep answers and grading material outside the runtime instructions. Follow the host's release and installation process.

Sources

The Agent Skills specification owns the package format. references/evaluation.md owns evaluation and names the upstream workflows it uses.

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.