repo-dependency-graph

# repo-dependency-graph Turn a skill repo (or a whole campaign package) into a clean, interactive **use-dependency graph**: read the original `SKILL.md` + design docs, reconstruct the real `use` edges, and render one offline HTML per repo that looks and behaves like an Obsidian graph (force-directed, draggable, neighbourhood-highlight, HTML hover tooltips). This skill exists because a repo's frontmatter alone gives a broken graph — real routing lives in prose, in `references/*-index.md`, and in same-layer escalation handoffs. Reconstructing the true graph takes a careful read of each repo, then a deterministic render. This skill captures both halves so the result is consistent every time and reusable across all repos. ## The model (read `references/layer-rules.md` for the full rules) - **5 vertex types**: `campaign` (red) · `strategy` (cyan) · `tactic` (yellow) · `sop` (purple) · `references` (gray dashed, = a `.py`/`.md` helper file). - **One edge type `use`**: `A -->|use| B` = A invokes/orchestrates B (caller → callee). - **Layer comes from frontmatter `type:` / `layer:`** (NOT `execution:`). Infer from body+README only if absent, and record the reasoning. - **Escalation = same-layer `sop → sop` use edge** (locked decision): "escalate to X" / "for deeper analysis use X" / "import X" are all drawn as `use` edges. Do NOT promote a skill's layer just to make the edge look legal — same-layer handoff is a first-class edge here. ## Workflow ### 1. Read the repo an...