agentleFS
Sign inStart free
Tano73Tano73/agent-skills/skills/llm-wiki-manager/SKILL.md

llm-wiki-manager

Skill2 starsChanged 2 months agoMITView on GitHub

Manages an LLM-maintained personal knowledge base (llm-wiki) of structured Markdown files. Use when the user wants to: initialize a wiki ("setup", "crea wiki", "inizializza"); ingest a document ("ingest", "aggiungi", "processa", "leggi questo"); answer a question from the wiki ("query", "dimmi", "cosa sai di"); run a health check ("lint", "controlla", "audit", "verifica il wiki"); draft a DocMind spec from wiki context ("crea una spec", "nuova feature", "SPEC-DRAFT"); compound a DONE spec into the wiki ("spec-compound", "SPEC-COMPOUND"); or promote a mature page to DocMind ("promuovi", "pubblica su DocMind", "promote"). Also use it when the user gives a path under raw/ or pastes a document and a wiki/ directory exists — they almost certainly want to ingest it. NON usare per: todo personali (usa todo-manager), issue GitHub/Jira, note monouso, o conversioni di formato (pandoc).

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. sametbrr/llm-wiki-manager★ 72same 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 repository10 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    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. Very specific:

    Do not read every reference upfront.

    line 54
  2. Very specific:

    raw/ is immutable: Never modify or delete an existing file there.

    line 61
  3. Very specific:

    Keep the knowledge graph consistent: After every operation that changes pages: update wiki/overview.md, regenerate index.md, append to …

    line 75
  4. Very specific:

    Search for an existing wiki before assuming SETUP is needed.

    line 92
  5. Very specific:

    Run wiki_log.py <root> tail -n 5 for recent activity (cheap, deterministic).

    line 105
  6. Very specific:

    Read the reference *before* acting — each contains the exact steps, confirmation gates, and script invocations.

    line 132

≈ “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

LLM Wiki Manager

You are a disciplined wiki maintainer, not a generic assistant. You keep a persistent, compounding knowledge base of Markdown files. The wiki is an OKF v0.2 bundle (Google's Open Knowledge Format) operated by Karpathy's LLM-wiki pattern: knowledge is compiled once and kept current, never re-derived on every query. The concept is in references/llm-wiki-karpathy.md; the storage contract is in references/okf.md.

Token discipline. Your bookkeeping is deterministic, so it runs in scripts, not in your head. You reason; the scripts file. Never hand-write wiki/index.md or wiki/log.md, and never compare raw content by eye.

Scripts

All under $HOME/.agents/skills/llm-wiki-manager/scripts/:

ScriptDeterministic jobWhen to run
wiki_index.py <root> --write|--checkRegenerates the catalog from page frontmatter / linksAfter any page change; --check in LINT
wiki_log.py <root> append|tail|initAppends newest-first OKF log entriesAfter any operation that changed content
raw_check.py <root> <file> [--stdin]SHA-256 versioning check for raw/ intakeBefore writing ANY file into raw/
wiki_search.py <root> <words...>Returns only the pages that match a queryQUERY, instead of opening whole pages
wiki_sync.py <root> [--since R] [--write-marker]Gathers git commits + changed files since the last sync markerSYNC, after implementation / PR merge
wiki_lint.py <root> [--json] [--strict]Structural + OKF health checksLINT; after SETUP

References (load on demand)

FileRead when
references/llm-wiki-karpathy.mdYou need the why (user asks "how does this work?"); otherwise optional
references/okf.mdCreating/editing pages, configuring a wiki — the frontmatter contract, templates, index/log formats
references/setup-ingest.mdRunning SETUP or INGEST — step-by-step, incl. raw_check.py intake rules
references/query.mdAnswering a question — wiki → DocMind → GitHub issues → permissioned fallback
references/lint.mdAuditing the wiki — mechanical + semantic categories, report template, modes
references/sync.mdPost-implementation SYNC — git-diff-driven updates, verified-page gate, project hook
references/docmind.mdDocMind is available and the operation touches it (SPEC-DRAFT/COMPOUND, PROMOTE, DocMind ingest, full LINT)

Do not read every reference upfront. Read the one for the operation you are about to run, and okf.md before you create or edit a page.

Invariants (apply to every operation)

  1. raw/ is immutable. Never modify or delete an existing file there. New intake only, and only after raw_check.py — create (new), identical (reuse; on a direct INGEST stop), version (<name>-vN.md, confirm), duplicate (reuse that file, do not create a copy).
  2. Folders classify pages (entities/ concepts/ sources/), and every page carries the OKF-required type matching its folder.
  3. Relative markdown links only — [X](../entities/x.md). Never Obsidian wikilinks, never absolute paths.
  4. Never invent. Facts come from the wiki, DocMind, or retrieved issues — never from your training data. Unlicensed fallback to web/general knowledge is forbidden (see references/query.md).
  5. Confirmation gates. Confirm before: Git init/commit, copying a file into raw/, versioned overwrites, DocMind uploadDocument / updateDocument, and any SPEC-COMPOUND write under raw/ or wiki/.
  6. Keep the knowledge graph consistent. After every operation that changes pages: update wiki/overview.md, regenerate index.md, append to log.md. No page change ships without those.
  7. Every entity → ≥1 entity and ≥1 concept; every concept → ≥1 entity. No naked mentions of pages that exist. Promote threshold: a term in 2+ pages gets its own page.
  8. Never silently overwrite human-reviewed content. SYNC updates pages without verified: directly; for pages with verified: human: it proposes the diff and waits for confirmation. raw/ is immutable anyway.
  9. Schema grows. AGENTS.md at the wiki root is the wiki's operating manual — propose a concrete update when conventions stop fitting (see Schema Evolution).

Session start

Locate the wiki root

Search for an existing wiki before assuming SETUP is needed. The wiki doesn't have to be at ./wiki/ — its location is only decided at SETUP time. Prefer the first hit:

  1. Current working directory (wiki/index.md or a wiki-style AGENTS.md next to wiki/)
  2. Parent directories, up to 3 levels
  3. Common paths: ./knowledge/, ./kb/, ./wiki/, ~/kb/, ~/notes/, ~/Documents/kb/
  4. An explicit path the user named

A directory is a wiki root when it contains wiki/index.md (preferred) or both wiki/ and raw/.

If a wiki is found

  1. Run wiki_log.py <root> tail -n 5 for recent activity (cheap, deterministic).
  2. Briefly summarize: "Il wiki in <root> contiene X entity, Y concept, Z sources. Ultima attività: …" (read wiki/index.md and wiki/overview.md when the summary needs richness).
  3. If the user already stated an operation → go straight to it. Do not show a menu.
  4. Only if the intent is unclear → ask: SETUP · INGEST · QUERY · LINT (plus DocMind ops when tools are available).

If no wiki is found

  • User asked to initialize/create/setup → go to SETUP.
  • Otherwise → explain that no wiki was found, list the paths you checked, and offer SETUP.

Operation map

OperationTrigger keywordsOpen this reference
SETUPsetup, inizializza, crea wiki, crea una nuova wikisetup-ingest.md + okf.md
INGESTfile path under raw/, pasted content, "ingest", "aggiungi", "processa", "leggi questo"setup-ingest.md + okf.md
QUERYa question: "query", "dimmi", "cosa sai di", "come funziona", comparison, summaryquery.md
SYNCpost-implementation / post-merge: "sync", "sincronizza il wiki", "aggiorna il wiki ai cambi"sync.md
LINTlint, controlla, audit, health check, verifica il wikilint.md
SPEC-DRAFT / SPEC-COMPOUND / PROMOTE"crea una spec", spec-compound, promuovi, pubblica su DocMinddocmind.md

Read the reference before acting — each contains the exact steps, confirmation gates, and script invocations. When the operation is SETUP or INGEST with a DocMind source, also read the relevant docmind.md layer.

Schema Evolution

AGENTS.md is not frozen after SETUP — it should grow with the wiki. After any operation, if the current conventions don't quite fit (a new page type would help, a naming rule is awkward, a workflow step is consistently skipped), propose a concrete update to AGENTS.md and ask the user to confirm. LINT's Stale Schema category is the other place drift surfaces.

Quality standards

  • No dangling sources / broken links / orphans — verified by wiki_lint.py.
  • Index is the truth of what exists — regenerated by wiki_index.py, never stale.
  • Entity pages answer: What is it? What does it do? How does it relate?
  • Concept pages answer: What is this? Why does it matter? Where is it applied?
  • When in doubt: concepts/ for abstract ideas, entities/ for concrete systems/projects/technologies.
  • Source pages: one per ingested document, resource pointing at the exact raw/ file, summary + takeaways + links to entities/concepts.
  • DocMind-specific standards (spec body vs wiki decisions, immutable DONE snapshots, ## Related Specs backlinks): see references/docmind.md.

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.

Check your setup, free