菜单
Set up and operate an LLM-maintained personal knowledge base wiki. Use when building a persistent, compounding knowledge base from source documents -- articles, papers, notes, transcripts, or any source material.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Create, review, audit, and list Architectural Decision Records for any software project. Trigger on: "create an ADR", "document this decision", "ADR for...", "what ADRs exist", "review this ADR", "update INDEX.md", architectural trade-offs, or design pivots affecting multiple layers or files. Project-agnostic -- works with any tech stack. Backed by Karpathy principles: think before deciding, simplest sufficient ADR, surgical scope.
Audit Codex skills for visibility flags, deterministic vs non-deterministic steps, and composability opportunities. Reads .Codex/skills/ (or a user-specified path), produces a punch list per skill, then shows proposed rewrites with a changelog. Trigger on: "audit my skills", "review my skill files", "which skills auto-fire?", "find duplicated logic in my skills", "should any skills be internal-only?", or any request to review, improve, or rationalize a set of SKILL.md files. Never modifies any file -- proposes changes and waits for approval.
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
Decision guide for delegating to caveman-style subagents. Tells the main thread WHEN to spawn `cavecrew-investigator` (locate code), `cavecrew-builder` (1-2 file edit), or `cavecrew-reviewer` (diff review) instead of doing the work inline or using vanilla `Explore`. Subagent output is caveman-compressed so the tool-result injected back into main context is ~60% smaller -- main context lasts longer across long sessions. Trigger: "delegate to subagent", "use cavecrew", "spawn investigator/builder/reviewer", "save context", "compressed agent output".
Ultra-compressed commit message generator. Cuts noise from commit messages while preserving intent and reasoning. Conventional Commits format. Subject ≤50 chars, body only when "why" isn't obvious. Use when user says "write a commit", "commit message", "generate commit", "/commit", or invokes /caveman-commit. Auto-triggers when staging changes.
Compress natural language memory files (AGENTS.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman-compress FILEPATH or "compress memory file"
| name | llm-wiki |
| version | 2.0 |
| category | utilities |
| execution_speed | slow |
| token_efficiency | low |
| triggers | ["personal knowledge base","build a persistent wiki","maintain knowledge base","compounding knowledge","source documents"] |
| cache_key | llm-wiki-2.0 |
| dependencies | [] |
| description | Set up and operate an LLM-maintained personal knowledge base wiki. Use when building a persistent, compounding knowledge base from source documents -- articles, papers, notes, transcripts, or any source material. |
| disable-model-invocation | true |
Build and maintain a persistent, compounding personal knowledge base. The LLM writes and updates the wiki; you provide sources and ask questions.
Instead of RAG (re-deriving from raw docs each query), the LLM builds a persistent wiki -- structured, interlinked markdown files. Knowledge compounds: cross-references, contradictions, and synthesis are built incrementally, not re-derived per query.
Key difference: The wiki is a persistent artifact. Cross-references already exist. Contradictions already flagged. Synthesis already reflects everything read. Gets richer with every source added.
You never write the wiki yourself. You curate sources, direct analysis, ask questions. The LLM does summarizing, cross-referencing, filing, bookkeeping.
| Layer | Owner | Description |
|---|---|---|
| Raw sources | Human | Immutable source docs -- articles, papers, images. Never modified by LLM. |
| The wiki | LLM | LLM-generated markdown: summaries, entity pages, concept pages, synthesis. |
| The schema | Human + LLM | Config doc (AGENTS.md/AGENTS.md) defining wiki conventions and workflows. |
docs/Obsidian Vault/[project name]/
wiki/
raw/ Raw source documents (human-curated, LLM reads only)
wiki/ LLM-generated pages
index.md Content catalog -- LLM reads first on every query
log.md Append-only chronological record of operations
entities/ Pages for named entities (people, places, orgs)
concepts/ Pages for ideas and themes
sources/ Summary pages per ingested source
Drop a new source into raw/ and say "ingest [filename]".
Flow:
Ask questions against the wiki. LLM searches index.md, reads relevant pages, synthesizes answer with citations.
File good answers back as new pages. A comparison you asked for, an analysis, a discovered connection -- these compound in the wiki.
Periodic health check. Ask: "lint the wiki".
LLM looks for:
index.md -- Content-oriented catalog. Every page listed with link + one-line summary + optional metadata. LLM updates on every ingest. LLM reads first on every query to find relevant pages.
log.md -- Chronological record. Append-only. Format each entry: ## [YYYY-MM-DD] operation | title. Parseable with grep.
The schema (AGENTS.md or AGENTS.md for the wiki repo) tells the LLM:
Co-evolve the schema with the LLM as you learn what works.
If using Obsidian as the IDE:
Humans abandon wikis because maintenance burden grows faster than value. LLMs don't get bored, don't forget to update cross-references, can touch 15 files in one pass. The wiki stays maintained because maintenance cost is near zero.
Human's job: curate sources, direct analysis, ask good questions, think about meaning. LLM's job: everything else.
/obsidian-vault -- Sets up Obsidian vault structure. Use before starting an LLM wiki with Obsidian as the IDE./memory-to-vault -- Converts Codex's persisted memory into a vault outline. Useful for seeding a new wiki with existing knowledge./grill-with-docs -- Creates CONTEXT.md and ADRs for a project. Feeds naturally into an LLM wiki as source material.