What problem does it solve? Understanding how a large repository is wired — what calls what, what connects to what — normally means reading hundreds of files. This Skill builds a static code graph of the repo with the graphify CLI, keeps it on the filesystem as a single greppable graph.json, and refreshes it automatically on every commit, so structural questions are answered by query instead of by manual exploration. ## Core Features & Use Cases - Graph building: Runs graphify . --no-viz --code-only to produce graph.json in graphify-out/ with no LLM API key required; docs, PDFs, and images are opt-in only. - Natural-language querying: Answers questions like "what connects the install scripts to the skills catalog?" via graphify query, plus graphify path and graphify explain for tracing routes and explaining nodes. - Commit-based freshness: Installs a Git hook and union merge driver so graph.json is rebuilt and merged on every commit, with a built_at_commit field as the single source of truth for staleness. - Codemap reconciliation: Ships scripts that merge graph-derived facts into human-written codemap notes under a KEEP/DERIVE split, reporting conflicts without overwriting human judgment. - Use case: Ask "how is the billing subsystem wired?" and get an answer from the graph, then fold the verified structure into .crew/codemap/ notes during an upgrade run. ## Quick Start Ask the assistant to build the code graph for this repository and then query what connects two subsystems you name.
