Produce a thorough architecture document grounded in the graph — the kind of brief a raw LLM can't write because it's built from the codebase's real clusters, real edges, and precomputed business summaries. Arguments: none (cwd graph; pass a graphId to target another). Needs summaries for the richest output — follow…
Render the graph visually. Argument: [architecture|cluster|flow|deps] (default architecture). For cluster/flow, an optional second arg is a seed node id/name.
Give a newcomer a real orientation to the codebase, scaled to its size and grounded in the graph. Argument: optional path to scope the explanation (file/folder); omitted = whole repo. Needs summaries for best results — follow the freshness/summarize-permission policy in SKILL.md first (structure-only still gives stack…
Compare dependency state between two points — two Git refs, two SBOM files, or two container image tags — and report what changed. diff runs the scan pipeline against both states, so the same target flags and gates apply to each side.
Answer "why is this package here?" — trace every dependency path that pulls a package into the graph. explain runs the same detect and match stages as scan, then walks the graph from the project roots to the package you named.
Resolve a target's dependencies into one consolidated graph, then optionally enrich, analyze, audit, and render it. scan is the pipeline every other graph command builds on — flags you learn here work the same way on explain and diff.