This generates a persistent knowledge base in `.nexus-map/` that lets any AI session cold-start on an unfamiliar codebase without fumbling around. It runs a five-stage PROBE protocol using Tree-sitter AST extraction and git forensics to produce `INDEX.md`, system boundaries, dependency graphs, test coverage maps, and change hotspots. Supports 30+ languages out of the box with hooks for custom extensions. The real value is in the discipline: it forces explicit evidence tracking, blocks you from writing "maybe" without stating what evidence is missing, and ensures `implemented` nodes actually point to real code paths. Requires Python 3.10+ and shell access. If you just need ad-hoc file queries mid-session, use nexus-query instead.
npx skills add https://github.com/haaaiawd/nexus-skills --skill nexus-mapper