plan

# /supergraph:plan Scan codebase, map blast radius, create machine-readable plan. Announce: "📐 /supergraph:plan — scanning codebase, creating plan..." ## Quick Gate < 20 lines, ≤2 files, no hub/bridge, complexity <10 → skip to `/supergraph:tdd`. ## Steps **0. Read CONTEXT.md (if exists):** ```bash cat CONTEXT.md 2>/dev/null | head -60 ``` Use domain vocabulary from CONTEXT.md in all plan task descriptions — never use raw file/class names where a domain term exists. **1. Read the codebase (MANDATORY before planning):** - Read config file → language, framework, versions - Read 2-3 source files near target area → naming, imports, error handling - Read 1-2 test files → test structure, assertion style **2. Ensure graph:** Reuse `/supergraph:scan` context. If not done → run scan first. Requires `CBM_PROJECT` + healthy `index_status`; if stale/degraded → `index_repository` (absolute path). **3. Graph analysis (parallel where independent):** See `references/codebase-memory-contract.md#Lifecycle`. Run `detect_changes`, `search_graph`, `trace_path` (inbound/outbound/data-flow), and `get_architecture` (overview/clusters/boundaries/hotspots) **in parallel** (no dependencies). After `get_graph_schema`, run recipes `hubs`, `bridges`, `test-gaps`, `cross-boundary` (requires schema). Derive risk from evidence; preserve escalation: >20 files STOP, hub/bridge needs approval. Respect `scan: TTL` — skip `index_repository` if fresh per scan logic. **3b. Serena (optional):** See `serena/...