agentleFS
keysersoosekeysersoose/claude-code-setup/skills/wiki-ingest/SKILL.md

wiki-ingest

Skill2 starsChanged 4 months agoMITView on GitHub

Ingest a source into the global LLM wiki (the knowledge base path defined in your global CLAUDE.md) — Karpathy's first LLM-wiki operation. Takes a URL, file, export, or pasted material; produces a cited wiki page, updates the index and related pages, and logs the ingest. Use when asked to "ingest this", "add to the wiki", "save this to the wiki", "wiki this", or when the user shares an article/report/export they want kept as knowledge.

45 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use wiki-status, wiki-capture and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • nexuschainsSkill

    wiki-status

    Reports the current state of the wiki vault — page counts, category breakdown, hub pages, orphans, stale entries, and a delta showing what has changed since the last ingest. Use this skill when the operator asks "what's the status", "wiki dashboard", "how many pages", "what's stale", "what's been ingested", "show me the delta", or "wiki health overview". Also surfaces structural insights (top hubs, fragmented clusters, bridge pages) when the operator asks "wiki insights", "what's central", or "show me the hubs". Read-only: this skill reports and recommends; it does not modify any files.

    Also in 5 projects that use this

  • jehrhardtSkill

    wiki-capture

    Quickly capture raw thoughts, links, images, or notes to today's Obsidian daily note from anywhere in Pi. Use when the user wants to jot something down for later wiki processing.

    Also in 6 projects that use this

  • Ivy-Chen1999Skill

    wiki-research

    Autonomously research a topic via multi-round web search, synthesize findings, and file structured results into the Obsidian wiki. Use this skill when the user says "/wiki-research [topic]", "research X", "find everything about Y", "do a deep dive on Z", "autonomous research on X", or wants comprehensive, web-sourced knowledge on a topic filed directly into their wiki.

    Also in 5 projects that use this

  • nexuschainsSkill

    cross-linker

    Scans the wiki vault and adds missing [[wikilinks]] between pages that are related but not yet connected. Use this skill when the operator says "link my pages", "find missing links", "cross-reference the wiki", "connect my wiki", "add wikilinks", or after any large ingest to weave new pages into the existing knowledge graph. Also trigger when orphaned pages need connecting or when wiki-lint reports fragmented tag clusters. This is a write-heavy skill — it modifies page files to insert links, unlike wiki-lint which only reports.

    Also in 4 projects that use this

  • Ivy-Chen1999Skill

    wiki-setup

    Initialize a new Obsidian wiki vault with the correct structure, special files, and configuration. Use this skill when the user wants to set up a new wiki from scratch, initialize the vault structure, create the .env file, or says things like "set up my wiki", "initialize obsidian", "create a new vault", "get started with the wiki". Also use when the user needs to reconfigure their existing vault or fix a broken setup.

    Also in 5 projects that use this

  • bozhouDevSkill

    wiki-init

    Use when bootstrapping a new personal wiki for any knowledge domain — research, codebase documentation, reading notes, competitive analysis, or any long-term knowledge accumulation project.

    Also in 7 projects that use this

  • +1 more from nexuschains
  • +3 more from Ivy-Chen1999

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 36 edited versions we read keep these.

  • 4

    Distinctive

    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:

    Read and discuss: Read the source fully (delegate to a cheap subagent if it's very long.

    line 26
  2. Very specific:

    Propagate: Update index.md if a new page was created (one line).

    line 35
  3. Distinctive:

    When a QUESTION the user asks produces a genuinely reusable synthesized answer (research, comparisons, decisions), offer to file it back as …

    line 43
  4. Distinctive:

    No secrets/credentials into the wiki, ever.

    line 47
  5. Distinctive:

    If the source contradicts the wiki, the WIKI yields only after you verify the source is right — otherwise note the conflict in the page.

    line 49

“Common” is measured: kept by at least 6 in 10 of the 36 edited versions of this file we could read. “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

/wiki-ingest — turn a source into compounding knowledge

WIKI = the global knowledge base path defined in your global CLAUDE.md. One source per ingest, supervised — Karpathy's own practice ("ingest one at a time, stay involved to guide what to emphasize"). Ask the user ONE question only if the target domain/project is genuinely ambiguous; otherwise infer and proceed.

Steps

  1. Capture the raw source.
    • URL → fetch it; save a pointer entry (URL + date + one-line what-it-is) in raw/README.md. Don't copy full web content into raw/.
    • Local file (PDF/export/doc) → leave it where it lives if large; add a pointer in raw/README.md. Small standalone files (<1MB) may be copied into raw/<topic>/.
    • Pasted text → save verbatim as raw/<topic>/<slug>.md with a source: frontmatter line.
    • raw/ is immutable: never edit anything already there.
  2. Read and discuss. Read the source fully (delegate to a cheap subagent if it's very long; verify the load-bearing claims yourself). Surface the 3-5 takeaways that matter FOR THE USER'S CONTEXT (their projects, their business) — not a generic summary.
  3. Write the wiki page. Decide placement:
    • Cross-project knowledge → domains/<domain>.md (extend existing page if one fits — update-don't-duplicate).
    • About a person/company → people/.
    • Project-specific → that project's page, or a subpage linked from it.
    • Genuinely new topic → new page in the best-fitting folder. Page format: frontmatter (type, created, updated, tags, source), one-line summary at top, the distilled knowledge, explicit citation link to the raw source/pointer, links to related pages.
  4. Propagate. Update index.md if a new page was created (one line). Touch the 2-5 RELATED pages whose content this source changes — add a line or fix a now-wrong claim (delete-when-wrong). Don't fan out further; lint catches the rest.
  5. Log + commit. Append to today's log: ## [YYYY-MM-DD] ingest | <domain/project> | <source title> with 2-3 bullets. Then git -C <WIKI> add -A && git commit -m "ingest: <source title>".
  6. Report. Tell the user: what was learned (the takeaways), where it landed, which pages were updated, and anything the source contradicted in the existing wiki.

Query write-back (bonus rule)

When a QUESTION the user asks produces a genuinely reusable synthesized answer (research, comparisons, decisions), offer to file it back as a wiki page via this same flow — Karpathy's "good answers can be filed back into the wiki as new pages."

Guardrails

  • No secrets/credentials into the wiki, ever.
  • Confidential or NDA-covered material: record only what you can evidence and keep it inward-facing — never let it leak into outward-facing work.
  • If the source contradicts the wiki, the WIKI yields only after you verify the source is right — otherwise note the conflict in the page.

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.