Answer questions and hold grounded conversations over the wiki by first mapping relevant pages, then reading them generously and synthesizing from the current state. Use when the user wants to ask, recall, discuss, compare, or reason about information already stored in the wiki.
Help the user think with their stored knowledge, not just retrieve snippets. Answers should be grounded in the current wiki and can be saved back into it when useful.
Assume README.md and AGENTS.md define the project model. This skill operationalizes the query side of that model.
Use a two-phase method optimized for recall.
Map the candidate set before deep reading.
Lean heavily on:
wq for frontmatter-driven discovery,rg for broad grep-style keyword and phrase matching,index.md as a quick catalog if it exists,log.md when recency or evolution matters.Prefer wq as the primary primitive and treat index.md as a convenience view, not the source of truth.
Start broad. Over-recall first. The spirit is "lots of context" rather than premature precision.
Once the relevant pages are mapped, read the selected pages generously and in full. Follow nearby pages, links, adjacent concepts, and recent history when they help the synthesis.
Selective traversal comes first; generous reading comes second.
If the needed information is not in the wiki yet, say so plainly and pivot into ingestion of new material.