From llm-wiki
Answer a question by searching and synthesizing across an LLM Wiki vault. Use when the user asks "what does the wiki say about…", "query the wiki", "ask the wiki", or any substantive question that should be answered from the vault's accumulated knowledge rather than from general training data. Substantive answers can be filed back as analysis pages so explorations compound.
How this skill is triggered — by the user, by Claude, or both
Slash command
/llm-wiki:wiki-queryThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Synthesize an answer from the wiki's contents. Good answers should be filable back into the vault so explorations compound over time.
Synthesize an answer from the wiki's contents. Good answers should be filable back into the vault so explorations compound over time.
Confirm the current directory is a wiki vault:
CLAUDE.md, wiki/index.md, wiki/log.md all exist.If any are missing, stop and suggest running wiki-bootstrap first.
Read CLAUDE.md before answering. It defines the page types, confidence scale, and conventions this vault uses. Citation format and filing rules come from there.
Start with wiki/index.md. Use its categories and one-line summaries to identify candidate pages. This is fast and works well up to a few hundred pages.
If the index is not enough (e.g. the question is about a concept not yet catalogued, or the vault has grown past ~200 pages), also grep across wiki/ for key terms from the question.
Pull 3–8 candidate pages initially. Expand only if coverage is clearly thin. Do not pre-fetch the whole vault.
Answer using only content present in the wiki and the raw sources it references. Cite everything:
[[source-YYYY-MM-DD-short-title]].[[page-name]].Confidence rules:
Gap handling — if the wiki can't answer the question, offer one or both of:
Match format to the question. Obsidian renders mermaid, tables, and callouts natively — use them when they clarify.
| Question shape | Recommended format |
|---|---|
| "What is X?" / "Tell me about Y" | Prose summary with wikilinks |
| "Compare X and Y" | Table |
| "What changed when?" | Timeline or dated list |
| "What depends on X?" / "What does X touch?" | List, or mermaid dependency graph |
| "Walk me through …" | Ordered steps |
| "Summarize everything about …" | Short narrative + bullet citations |
If the answer synthesizes multiple pages or contains reasoning that would be costly to re-derive, offer to file it:
File this as an analysis page? (y/n)
If yes:
wiki/analysis/analysis-YYYY-MM-DD-short-title.md with full frontmatter (type: analysis, confidence matching the weakest underlying source).wiki/index.md under the Analysis category.wiki/log.md:## [YYYY-MM-DD] query | Brief Title
Filed: [[analysis-YYYY-MM-DD-short-title]]
From: [[page-a]], [[page-b]], [[page-c]]
If no or the answer is a simple lookup: append a lighter log entry:
## [YYYY-MM-DD] query | Brief Title
Answered from: [[page-a]], [[page-b]]
wiki-ingest's job. If you notice the wiki is missing a source the user has, tell them.wiki-lint; do not silently edit.npx claudepluginhub bmentges/knowledge-base-template --plugin llm-wikiFetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
Applies a firm's KYC/AML rules grid to parsed onboarding records: assigns risk rating, checks required documents, outputs rule outcomes with citations, and routes for escalation.
Generates daily or weekly digests of activity from connected sources (chat, email, docs, tasks, CRM), highlighting action items, decisions, mentions, and project updates.