explore

Explore

Goal: understand the area from real source before touching it. Prefer the mimirs tools over grep + reading whole files.

1. Where is it — search "<topic>" for ranked files, then read_relevant "<topic>" to pull the actual functions/sections with line ranges (not just paths). Two chunks from one file can both appear.
2. How it connects — project_map(focus: <key file>) for its neighborhood; dependents <file> for who relies on it; depends_on <file> for what it pulls in.
3. Read the core — read_relevant "<specific behavior>" on the key files; follow the helper calls a named file delegates to.
4. History & why — git_context for what changed recently; file_history <file> for how a file evolved; search_commits "<topic>" for why past decisions were made.
5. Prior context — search_conversation "<topic>" and search_checkpoints "<topic>" for earlier discussion and decisions, so you don't re-derive what's already known.
6. Caveats — get_annotations on the key files for known bugs, fragile spots, or constraints before trusting the code.

Finish: summarize in plain language — entry points, key files, data flow, gotchas — each with a file:line citation. Flag anything still unclear instead of guessing.