cortex

Run the 100-question recall eval against cortex's current index to catch recall regressions. Use when the user asks "run the eval", "test cortex recall", "evaluate recall quality", after changes to the write/recall/ingest pipelines, after tuning activation/visibility/PPR parameters, or when investigating whether a recall change helped or hurt. The harness is a black-box semantic eval: it runs `cortex recall` against 100 factual codebase questions and scores top-5 results by keyword hit-rate.

Cortex is YOUR persistent memory across sessions in this repo. Use to recall prior observations before any non-trivial task and to record decisions, root causes, and discoveries during work. MUST consult on every task that touches code, architecture, design decisions, bug investigations, performance work, or anything the user asks "how do we", "why did we", "what did we decide", "do we already". MUST write observations at every natural checkpoint (root cause found, decision made, bug fixed, benchmark run, surprising behavior discovered).

Use when the user needs to run GitNexus CLI commands like analyze/index a repo, check status, clean the index, generate a wiki, or list indexed repos. Examples: "Index this repo", "Reanalyze the codebase", "Generate a wiki"