agentleFS
Sign inStart free
MicrosoftKnown publishermicrosoft/amplifier-bundle-work/skills/skill-creator/SKILL.md

skill-creator

Skill0 starsChanged 27 days agoMITView on GitHub

Create or update Amplifier skills with concise routing, portable resources, and actual loader validation. Use for reusable workflows and skill authoring.

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 183 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
  1. Very specific:

    Use a unique lowercase hyphenated name and a concise description explaining when to load the skill.

    line 15
  2. Very specific:

    Keep interactive work inline.

    line 23
  3. Very specific:

    Put essential steps and completion evidence in SKILL.md.

    line 28
  4. Distinctive:

    Test descriptions with ordinary requests and near misses using the full catalog.

    line 39

“Common” is measured: kept by at least 6 in 10 of the 183 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

Skill Creator

Inspect the destination's conventions and existing skill before editing. When available, load skills-assist for current Amplifier authoring guidance. Derive the workflow from concrete user tasks; include instructions that change useful decisions rather than repeating generic assistant behavior.

Use a unique lowercase hyphenated name and a concise description explaining when to load the skill. Add user-invocable: true when a slash command is useful. Match the name to its directory, use at most 64 characters with no consecutive hyphens, and keep the description within the standard's 1024-character limit and the destination's tighter routing budget. Use string compatibility (up to 500 characters) for actual environment requirements; keep metadata values strings. version, context, and user-invocable are host extensions, not standard Agent Skills fields. See https://agentskills.io/specification. Keep interactive work inline. Use context: fork only for a bounded independent task, and only with the current host's support. Fork allowed-tools uses module IDs; omit it when the full parent tool surface is appropriate. Do not pin a provider/model unless required by the user's requested capability.

Put essential steps and completion evidence in SKILL.md. Store conditional detail in linked references, deterministic helpers in scripts, and reusable output inputs in assets. Resolve resources relative to skill_directory or ${SKILL_DIR}; do not assume the caller's working directory.

Save to the authorized bundle or .amplifier/skills/<name> destination. Preserve unrelated files and notices. Parse frontmatter, verify linked resources, load through the actual load_skill tool, and test a realistic scenario including missing optional capabilities. Report loader, execution, and visual checks separately. Update an existing skill in place only when that target is requested.

Test descriptions with ordinary requests and near misses using the full catalog; explicitly naming a skill does not test selection. Record actual loads and independent output checks in fresh sessions. For a quality-improvement claim, compare against the previous version or a no-skill baseline with the same tools and model. See https://agentskills.io/skill-creation/evaluating-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.

Reviews can't be read right now.

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.

Check your setup, free