agentleFS
ViaknoxViaknox/tune-up/skills/skill-index/SKILL.md

skill-index

Skill0 starsChanged 22 days agoMITView on GitHub

Use when starting a multi-step job, when the user names a procedure, or when a keyword might match a stored skill. Retrieve one skill body. Do not load the index into the standing prompt.

4 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions4 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 4

    1. listener-He/java-harness-agent★ 16same name
    2. NiBorracho/skill-index★ 1same name
    3. ivolentjesrijnijssel-oss/Claude-skills★ 0same name
    4. ragnos-labs/skill-index★ 0same 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 repository3 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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:

    Add a location to an existing listing only if it is the same skill: tied to the same repo, or a fork of that repo.

    line 45
  2. Distinctive:

    If two listings still match, load neither.

    line 58
  3. Distinctive:

    Before saying done, score the success line.

    line 60
  4. Distinctive:

    If the store is down, use a local thin cache of listings only.

    line 61
  5. Very specific:

    Not in the library. Search the default list in marketplaces.json, plus any location the user added.

    line 74

≈ “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

Skill index

Part of Tune-Up. Tune-Up keeps a skill index so each bot loads one skill only when the job needs it.

The standing prompt keeps identity, lanes, and locks. Procedures live in an index and load only after a real match.

Index fields (thin listing)

The plugin database has one listing per skill. Match listings, not every copy.

  • name
  • purpose (when to use, one line)
  • keywords (exact hits)
  • do-not-use-when (hard veto)
  • success check (worked if… / fail if…)
  • where it may load (this chat, named rooms, never a group)
  • repo (the skill's source repo)

A listing does not store the body.

Locations

Each listing has one or more locations. A location is where that same skill was found.

  • store (files, Notion, a marketplace, a repo)
  • path or URL
  • repo, or the repo it was forked from
  • license, if known
  • last checked

Add a location to an existing listing only if it is the same skill: tied to the same repo, or a fork of that repo. Same name is not enough. Do not create a second listing for another copy of that same skill.

Overlapping purpose is not a merge. A different repo that is not a fork of this one is a different listing, even if the name matches.

Linked, not on the standing row:

  • Edge cases: trigger, do or don't, hard stop or warning
  • Lessons: date, what happened, skill that fired or should have, what changed, still open

Load order

  1. Match purpose and keywords on listings, against the current turn.
  2. Drop any hit whose do-not-use-when or edge case matches this turn.
  3. If two listings still match, load neither. Name both and why. A closer keyword does not cancel a second match. Extra locations on one listing are not a second match.
  4. Load one body from one location, the matching edge cases, and only lessons tied to this skill or this kind of miss. Prefer a confirmed license, then a local copy. Do not load every location.
  5. Before saying done, score the success line. If it cannot be scored from this turn, the score is not a hit. If it fails, write a lesson in the same turn.
  6. If the store is down, use a local thin cache of listings only. No body. No guess. Locks still hold.

Stores

Same fields, any store: files, Notion, Airtable, Coda, a sheet, or a marketplace. The store is what a human edits, or a source you import from. Keep a small local copy of the listing fields. Do not query every store on every message.

What never goes in the index

Standing locks, secrets, and another bot's private instructions.

Three flows

  1. In the library. If one listing matches the job, use it. The user does not need to ask. Do not search marketplaces.
  2. Not in the library. Search the default list in marketplaces.json, plus any location the user added. Use each source's API. Do not scrape. Show the relevant skills and ask if one can be added. Do not write until they say add.
  3. Named site. If the user names a site to search, search that site. Then ask if that location should be added to the default search list. Add it only if they say yes. A new search location does not fill the library.

Same repo, or a fork of that repo, is another link on the one listing. Do not guess a body.


Tune-Up, by Viaknox™. Code and text MIT licensed; brand art is not. Viaknox and Agent Tune-Up are trademarks of Viaknox Holding LLC; see TRADEMARK.md. Not affiliated with or endorsed by SpaceXAI (xAI).

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.