agentleFS
camrontaylorcamrontaylor/AI-OS-Template/.claude/skills/meta-find-skills/SKILL.md

meta-find-skills

Skill0 starsChanged 47 days agoMITView on GitHub

Choose the smallest useful set of installed AI-OS skills for a task. Use for find a skill, what skill handles this, can AI-OS do this, or a multifaceted task with uncertain routing. Scans the shared inventory and only the active client's skills, respects hidden skills, and names real capability gaps. Does not build skills, install tools or change accounts.

2 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

11 rules. About 0 are common, 5 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never read local client settings, context or another user's files in Team mode.

    line 25
  2. Distinctive:

    If the snapshot lacks a scoped inventory, report that limit rather than inferring client access from a path.

    line 26
  3. Very specific:

    Read the frontmatter of every direct child SKILL.md in the shared source, excluding underscore-prefixed support folders.

    line 38
  4. Very specific:

    Use .claude/skills/ as authority.

    line 43
  5. Very specific:

    Do not route this task back to meta-find-skills itself.

    line 47
  6. Distinctive:

    Use one skill for a focused task.

    line 52
  7. Very specific:

    Read each selected SKILL.md, its applicable local overrides, Context Needs and matching learnings before execution.

    line 56
  8. Very specific:

    Check the optional catalog only after installed coverage.

    line 63
  9. Very specific:

    If this install has a skills-library/INDEX.md, read only relevant parked entries.

    line 67
  10. Distinctive:

    Never install skills globally or replace local overrides.

    line 77
  11. Distinctive:

    Do not require usage-ranking scripts, a library, a connector or a paid service that this template does not ship.

    line 79

≈ “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

Find Skills

Choose one to three non-overlapping skills that cover the requested outcome. Complete the original task after routing; a list of skills alone only finishes a request whose outcome was the list itself.

Context Needs

SourceLoadPurpose
Root .claude/skills/*/SKILL.mdAll frontmatter; full selected skillsCanonical shared inventory
Active client .claude/skills/*/SKILL.mdFrontmatter; full selected skillsClient-only methods and same-name overrides
Active client's .claude/settings.local.jsonskillOverrides onlyExclude skills marked off in Solo mode
Selected SKILL.local.mdFull, root and active client when presentPreserve user-owned changes
Active context/learnings.md## meta-find-skillsPrior routing lessons, Solo only
_catalog/catalog.jsonTargeted after installed coverageOptional methods that can be restored

Resolve root and active scope through docs/agent-startup.md. In Team mode, read only shared methodology and the authorized injected inventory and preferences. Never read local client settings, context or another user's files in Team mode. If the snapshot lacks a scoped inventory, report that limit rather than inferring client access from a path. Respect the existing conversation-only boundary.

1. Identify the needs

Name the outcome and the capabilities it needs. Separate needs that change the correctness or usefulness of the result from optional extras. For multi-part work, map its phases before choosing a skill. Do not collect skills for minor needs the agent can already handle.

2. Read the complete installed inventory

Read the frontmatter of every direct child SKILL.md in the shared source, excluding underscore-prefixed support folders. Include only the active client's authorized local skills. Deduplicate inherited links by resolved path. A client's same-name skill takes precedence; an off override excludes that skill.

Use .claude/skills/ as authority. .agents/skills/ is its discovery adapter; catalogs and the current tool's picker may be stale. Use the public method's triggers, exclusions, Context Needs and available fallback to judge fit.

Do not route this task back to meta-find-skills itself. A runtime-only plugin may supplement a method, but label it runtime-only and do not promise portability.

3. Commit to the smallest route

Use one skill for a focused task. Choose two or three only when each covers a different material need. Prefer one exact-fit method over overlapping specialists. Order them by dependency, such as research, decision, then production.

Read each selected SKILL.md, its applicable local overrides, Context Needs and matching learnings before execution. Retain the original objective and accepted constraints when switching methods. Tool references describe a capability; map them to the current agent's available tools and documented fallback.

4. Handle a real gap

Check the optional catalog only after installed coverage. Restore a named optional skill through scripts/add-skill.sh when the user has asked for its installation; listing it does not mean it is installed.

If this install has a skills-library/INDEX.md, read only relevant parked entries. Parked sources are candidates for review, not live methods. If neither the catalog nor a local library covers the need, name the gap and proceed with available tools when useful. Offer to find or build a method only when it would improve the job.

External discovery is for a material uncovered need or an explicit request. Use current official source repositories and documentation, check scope and license, and keep any downloaded candidate under a task folder in the install's ignored .tmp/ directory. Report it as task-scoped. It becomes a shared skill only through the registration process in docs/building-skills.md and the user's decision. Never install skills globally or replace local overrides.

Do not require usage-ranking scripts, a library, a connector or a paid service that this template does not ship. Their absence is not a broken routing system.

Output

For a skill-finding request, give one chosen route with each skill's job and any material gap. For an execution request, name the selected method briefly and do the work. Keep rejected candidates out of the reply unless the choice turns on their tradeoff. Ask only when a missing fact changes the route.

Verification

  • An installed match is chosen without an external search or installation.
  • A multi-part request gets at most three distinct methods covering its needs.
  • A hidden or unauthorized client skill is excluded; sibling context is unread.
  • An optional or parked candidate is labelled accurately and stays uninstalled.
  • A missing optional service uses the documented fallback or a stated limit.
  • Selected methods are read fully, and the original requested outcome is checked.

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.