graph-querylisted

# Graph Query Ask the knowledge graph a relationship question and get a **cited answer grounded in the graph** — not a guess. The graph is built by `/graph` (graphify) over code + the committed `wiki/`. This skill only reads it. ## When to run - `graphify-out/graph.json` exists, and the user has a question about how things relate: - "what connects the intake flow to the billing module?" - "trace how a `/wiki` capture becomes a committed entry." - "explain the WAT execution model from the graph." - "shortest path between `wiki_lint.py` and the commit gate." - Use this for *relationships and explanations*. To (re)build or refresh the graph, run `/graph`. ## The output (always this shape) ``` ## Graph answer — <the question> **Answer** — the relationship/explanation, in plain sentences, grounded only in graph edges. **Path / nodes used** - <node A> —[<relation>]→ <node B> (EXTRACTED · src: <path:line>) - <node B> —[<relation>]→ <node C> (INFERRED — not literal in source) **Citations** — source_location for each EXTRACTED edge (path:line). **Gaps** — anything the question asked that the graph does NOT contain (say so plainly). ``` Every claim traces to an edge. If the graph doesn't have it, the answer says so — it never fills the gap from memory. ## Process 1. **Check the graph exists.** If `graphify-out/graph.json` is missing, stop and tell the user: "No graph yet — run `/graph` first to build it." Don't answer from training data or by re-reading files. 2. **P