AugmentClaude

LLM Wiki Manager

Convert web, GitHub, and YouTube URLs into a structured, citable knowledge base.

Installation

  1. Make sure Claude is on your device and in your terminal.

    Skills load from ~/.claude/skills/ when Claude Code starts up — so you need it on your machine first. If you don't have it yet, install it once with the command below, then run claude in any terminal to verify.

    One-time setup
    npm i -g @anthropic-ai/claude-code

    Already have it? Skip ahead.

  2. Paste into Claude Code or into your terminal.

    This copies the whole skill folder into ~/.claude/skills/pin-llm-wiki-ndjordjevic/ — the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Safe default: works for every skill.

    Faster alternative (instruction-only skills)

    Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates — they won't be downloaded and the skill will fail when it tries to load them.

    Quick install (SKILL.md only)
    Sign up to copy
  3. Restart Claude Code.

    Quit and reopen Claude Code (or any other agent that loads from ~/.claude/skills/). New skills are picked up on startup.

  4. Just ask Claude.

    Skills auto-activate when your request matches the skill's description — no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the “What this skill does” section above.

Prefer to read the source first? Open on GitHub.

When Claude uses it

Automates the Karpathy LLM Wiki workflow: turns web, GitHub, and YouTube URLs into well-structured, citable, wikilinked pages with automatic linting and sourcing — invoke with /pin-llm-wiki

What this skill does

/pin-llm-wiki

Automates Karpathy's LLM wiki pattern: drop URLs in inbox.md, the skill fetches, ingests, and maintains a cited, agent-readable knowledge base.

Trigger phrases

  • /pin-llm-wiki (with subcommands: init, ingest, lint, queue, remove)
  • “Pin this YouTube video to my LLM wiki”
  • “Ingest these research links into my wiki”
  • “Run pin-llm-wiki lint on this knowledge base”
  • “Queue these URLs, then batch-process the inbox”
  • “Initialize a new Karpathy-style wiki with pin-llm-wiki”
inbox.md (human drops URLs)
    ↓  fetch
raw/  (immutable source captures)
    ↓  ingest
wiki/  (LLM-maintained, cited, linked)
    ↓  lint
a healthy, queryable knowledge base

Phase 1 subcommands

CommandStatus
initimplemented
ingest [<url>]implemented (single-URL form auto-queues if URL is not already in inbox)
lintimplemented
remove <slug>implemented
queue <url> [<url> ...]implemented

Skill directory

This SKILL.md and all sibling files (ingest.md, init.md, lint.md, remove.md, queue.md, ingest-protocol.md, templates/...) live inside the skill directory: ~/.claude/skills/pin-llm-wiki/, ~/.copilot/skills/pin-llm-wiki/, ~/.cursor/skills/pin-llm-wiki/, or the project-local .claude/skills/ / .copilot/skills/ / .cursor/skills/ equivalents. In this repository the canonical copy is skills/pin-llm-wiki/. All templates/... and sibling-file paths in this skill are relative to whichever skill directory the loading tool used.

Dispatch

  1. Identify the subcommand from the invocation args (the first word after /pin-llm-wiki).
  2. Route — read the sibling file in this skill directory and follow its instructions exactly:
    • initinit.md (no Guard required — it scaffolds the wiki)
    • ingestingest.md (with or without a URL arg)
    • lintlint.md
    • removeremove.md
    • queuequeue.md
  3. Guard (every subcommand except init): confirm .pin-llm-wiki.yml exists in the current working directory. If absent, stop with: "No wiki found here (.pin-llm-wiki.yml missing). Run /pin-llm-wiki init to scaffold one first." Subcommand files repeat this check by reference; you only need to enforce it once per invocation.
  4. Do not proceed beyond this dispatch step before reading the target file.

Git policy (canonical)

Never run git commit or git push after any subcommandinit, ingest, refresh, lint, remove, queue, or any auto-fix — unless the human explicitly asked to commit in this conversation. Subcommand files reference this policy without restating it. The wiki's own AGENTS.md carries the same rule for downstream agents.

Related skills