wiki-graphlisted

# wiki-graph Structured graph queries over the wiki — **without a persistent graph store**. The wiki already has a knowledge graph: every `[[wikilink]]` is an edge, every page is a node, every frontmatter field is a property. This skill just computes over that graph on demand. Each invocation scans the markdown files, builds an in-memory graph, runs the query, prints the answer, and exits. Nothing is cached — the files are the source of truth, always. > Karpathy's wiki treats the filesystem as the database. `wiki-graph` treats the > filesystem as the graph. ## When to use - Structured Dataview-style questions: _"all `type: model` pages tagged `transformer` updated since March"_ - Traversal questions: _"what's within 2 hops of `[[claude-3]]`?"_ - Comparison discovery: _"what's the shortest path between `[[gpt-4]]` and `[[llama-3]]` — do any concepts bridge them?"_ - Topology questions: _"which pages are hubs? which are orphans?"_ - Visualizing a subgraph: render a Mermaid diagram of a cluster or neighborhood `wiki-search` is for **"find pages about X"** (ranked text relevance). `wiki-graph` is for **"pages where property P holds"** and **"pages connected to X"** (structural queries). ## Implementation The graph algorithm is implemented in `plugin/scripts/graph_query.py`. **The script is the source of truth; the prose below explains its behavior.** Don't reimplement the BFS, hub scoring, or tier filtering by reading files yourself — shell out to the script and forma