agentleFS
B-AI-botB-AI-bot/agent-maintained-vault/skills/vault-wiki/SKILL.md

vault-wiki

Skill0 starsChanged 18 days agoMITView on GitHub

Maintain a workspace of agent-maintained vaults (LLM-wiki pattern) — route inbox captures to the right vault, ingest sources into wiki pages, answer from the wiki, run the nightly lint, and propose every canon change as a pull request approved by a human. Use for any curator session, any ingestion, any lint, and whenever an agent wants to change an AGENTS.md, a CANON-* page, a SKILL.md or vaults.toml.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. alycd/agent-skills★ 0same name
    2. olafgeibig/skills★ 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 repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    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:

    Put the change on a branch, open a pull request whose body gives the reason, the source (a raw/ path or a URL) and any contradiction found.

    line 25
  2. Very specific:

    Write or update the wiki/ pages it touches: summary plus a link to the source.

    line 37
  3. Very specific:

    Read index.md, then the pages.

    line 46
  4. Distinctive:

    Write a secret · rewrite a raw file · put content in a vault that is not its own · edit a canon without a PR · merge a PR.

    line 64

≈ “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

Vaults are wikis kept by agents; canons are approved by humans

Mental model

The workspace holds several vaults, each with its own repository and access list (registry: vaults.toml, map: the workspace AGENTS.md). The owner sees everything; a partner or a client sees only their vault. So picking the vault is a confidentiality decision: client material in the company vault leaks to partners; personal material outside the owner vault leaks to everyone.

Each vault has three layers:

LayerWhereRule
Rawraw/immutable sources: add, never rewrite
Wikiwiki/pages written by agents; index.md catalogues them, log.md records every operation
Schema / canonAGENTS.md, CLAUDE.md, wiki/canon/, skills/*/SKILL.md, vaults.tomlpull request + human approval

Changing a canon: propose, don't edit

Put the change on a branch, open a pull request whose body gives the reason, the source (a raw/ path or a URL) and any contradiction found. One PR, one topic. Never present a value from an open PR as true. Only a human merges. Branch protection plus a CODEOWNERS entry on canon paths makes this enforceable, not just polite.

Ingest (inbox → right vault)

For each item in the owner inbox:

  1. Decide the vault: named client → that client's vault; company matter → company; personal → owner; doubt → leave it in the inbox and say so. An item that mixes scopes is split, each part to its vault.
  2. Move the source as-is into that vault's raw/ (content untouched).
  3. Write or update the wiki/ pages it touches: summary plus a link to the source.
  4. Contradiction with a canon → PR on the canon, never a direct edit.
  5. One line in index.md per new page; one entry ## [YYYY-MM-DD] ingest | <title> in log.md (vaultctl log <vault> ingest "<title>" --body "...").

In unattended runs: items older than 7 days only, at most 5 per run.

Query

Read index.md, then the pages. Quote canons for volatile facts. An answer that will be asked again becomes a wiki page (and a line in index.md).

Lint (nightly)

  1. vaultctl check <workspace> --since <yesterday's commit>: structure, canon frontmatter and review dates, broken links, pages missing from index.md, malformed log.md, secrets, private terms leaking across vaults, raw/ files modified or deleted.
  2. Judgement pass the script cannot do: wiki pages contradicting a canon, volatile values copied outside the canon, inbox items older than 7 days, PRs open for more than 3 days.
  3. A deleted raw file may be an intact move: compare contents before calling it a loss.
  4. Closed-without-merge PRs and review comments become feedback-* notes in the agents' memory: human corrections are how agents improve.

Output: a short message to the owner with counts per vault and PR links. Nothing to report = one line.

Never

Write a secret · rewrite a raw file · put content in a vault that is not its own · edit a canon without a PR · merge a PR.

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.