From rock-star-skills
Search the LLM Wiki using qmd (hybrid BM25/vector search). Finds relevant wiki pages, reads them, and synthesizes an answer with citations. Use when asking questions against the knowledge base.
npx claudepluginhub marvec/rock-star-skills --plugin rock-star-skillsThis skill is limited to using the following tools:
Search the wiki and synthesize an answer from the knowledge base.
Search and query the wiki — get cited answers, with automatic research when knowledge is missing. Three depth modes. Use when: 'check wiki', 'what do we know about', 'look up', 'find in wiki', 'search wiki', 'search for', 'find information about', 'do we have notes on', 'wiki context', 'existing knowledge', 'research', 'look into', 'find out about', 'investigate', 'what is', 'how does X work'.
Answers domain questions by reading and synthesizing project wiki pages, citing sources with wikilinks. Flags gaps and contradictions.
Answers questions from wiki vault by searching with scripts, citing claims via [[wikilink]]s, and synthesizing new MD pages for complex topics.
Share bugs, ideas, or general feedback.
Search the wiki and synthesize an answer from the knowledge base.
The user's argument is the search query. If no argument is provided, ask the user what they'd like to search for.
Use a multi-pronged search approach:
Use the qmd MCP tools to search the wiki collection:
{
"searches": [
{ "type": "lex", "query": "<key terms extracted from query>" },
{ "type": "vec", "query": "<full natural language query>" }
],
"collections": ["wiki"],
"intent": "<what the user is looking for>",
"limit": 15,
"minScore": 0.3
}
For complex queries, add a hyde sub-query with a hypothetical answer (50-100 words).
Read wiki/index.md and scan for relevant page titles and descriptions. This catches pages that qmd might miss due to indexing lag.
If the query contains specific terms, names, or identifiers, use Grep to find exact matches across wiki pages.
From the search results, identify the most relevant pages (typically 3-10). Read them in full using the Read tool or qmd get.
For each page, note:
If the initial pages reference other wiki pages that seem highly relevant, follow those links too (one level deep).
Compose an answer that:
[[Page Title]]Format the answer in clean markdown. Use headers, bullet points, and tables where they aid clarity.
If the answer is substantial and adds value beyond what's already in the wiki, offer to file it:
"This analysis could be saved as a wiki page. Would you like me to file it in
wiki/comparisons/orwiki/synthesis/?"
If the user agrees, write the page following standard wiki conventions (frontmatter, wikilinks, proper filename) and update the index.
Append to wiki/log.md:
## [YYYY-MM-DD] query | <short query description>
- Query: <full query>
- Pages consulted: [[Page1]], [[Page2]], ...
- Filed as: wiki/path/to/new-page.md (if applicable)
updated date in frontmatter. If a page hasn't been updated recently, mention this.