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
- google/capsemtheir own version, first seen Oct 30, 2026
- shreyas-makes/agent-skillstheir own version, first seen Jul 15, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- google/capsem★ 179same name
- shreyas-makes/agent-skills★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository96 other files it gives its agents.
- AGENTS.mdAGENTS.md
- team_context/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- fin-company-cfoSkill
- fin-invoice-reconciliationSkill
- fin-month-end-reportingSkill
- fin-personal-cfoSkill
- meta-memory-recallSkill
- meta-memory-writeSkill
- meta-skill-creatorSkill
- meta-skillifySkill
- meta-synthesize-localsSkill
- meta-toolifySkill
- meta-wrap-upSkill
- mkt-ab-testingSkill
- mkt-ad-creativeSkill
- mkt-adsSkill
- mkt-asoSkill
- mkt-brand-voiceSkill
- mkt-churn-preventionSkill
- mkt-cold-emailSkill
- mkt-co-marketingSkill
- mkt-community-marketingSkill
- mkt-content-repurposingSkill
- mkt-copy-editingSkill
- mkt-copywritingSkill
- mkt-croSkill
- mkt-directory-submissionsSkill
- mkt-emailsSkill
- mkt-eventsSkill
- mkt-free-toolsSkill
- mkt-icpSkill
- mkt-influencer-marketingSkill
- mkt-jab-hookSkill
- mkt-launchSkill
- mkt-lead-magnetsSkill
- mkt-marketing-loopsSkill
- mkt-offersSkill
- mkt-onboardingSkill
- mkt-paywallsSkill
- mkt-popupsSkill
- mkt-positioningSkill
- mkt-programmatic-seoSkill
- mkt-prospectingSkill
- mkt-public-relationsSkill
- mkt-referralsSkill
- mkt-sales-enablementSkill
- mkt-schemaSkill
- mkt-signupSkill
- mkt-smsSkill
- mkt-socialSkill
- mkt-ugc-scriptsSkill
- mkt-visual-identitySkill
- ops-analyticsSkill
- ops-attributionSkill
- ops-cronSkill
- ops-ingestSkill
- ops-loopifySkill
- ops-project-managementSkill
- ops-revopsSkill
- The first 60 shown.
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
- Distinctive: line 25
Never read local client settings, context or another user's files in Team mode.
- Distinctive: line 26
If the snapshot lacks a scoped inventory, report that limit rather than inferring client access from a path.
- Very specific: line 38
Read the frontmatter of every direct child SKILL.md in the shared source, excluding underscore-prefixed support folders.
- Very specific: line 43
Use .claude/skills/ as authority.
- Very specific: line 47
Do not route this task back to meta-find-skills itself.
- Distinctive: line 52
Use one skill for a focused task.
- Very specific: line 56
Read each selected SKILL.md, its applicable local overrides, Context Needs and matching learnings before execution.
- Very specific: line 63
Check the optional catalog only after installed coverage.
- Very specific: line 67
If this install has a skills-library/INDEX.md, read only relevant parked entries.
- Distinctive: line 77
Never install skills globally or replace local overrides.
- Distinctive: line 79
Do not require usage-ranking scripts, a library, a connector or a paid service that this template does not ship.
≈ “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.
Files
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
| Source | Load | Purpose |
Root .claude/skills/*/SKILL.md | All frontmatter; full selected skills | Canonical shared inventory |
Active client .claude/skills/*/SKILL.md | Frontmatter; full selected skills | Client-only methods and same-name overrides |
Active client's .claude/settings.local.json | skillOverrides only | Exclude skills marked off in Solo mode |
Selected SKILL.local.md | Full, root and active client when present | Preserve user-owned changes |
Active context/learnings.md | ## meta-find-skills | Prior routing lessons, Solo only |
_catalog/catalog.json | Targeted after installed coverage | Optional 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.
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
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.

