domain-extractlisted

# Domain Extract Analyze existing project sources to propose domain knowledge rules for approval. ## Dispatch Based on $ARGUMENTS after "domain": - `domain extract` → Step 1-5 (full extraction) - `domain list` → Step 6 (list only) - `domain sync-vault` → Step 7 (vault sync only) --- ## Step 1: Gather sources Read these project sources (skip any that don't exist): 1. **CLAUDE.md** — sections below `<!-- forge:custom -->` (user-written content) 2. **Auto-memory** — resolve the project's memory directory: - Compute project path hash: the memory lives in `~/.claude/projects/-<path-with-dashes>/memory/` - Read MEMORY.md index, then read each linked memory file - Filter: keep only entries with business/domain content (not purely technical) 3. **CLAUDE_ERRORS.md** — read all entries. Flag errors where the root cause is "Claude didn't know X about the business" 4. **Agent memory** — read `.claude/agent-memory/implementer/*.md` and `.claude/agent-memory/architect/*.md` - Filter: keep domain-relevant learnings (business logic, API behavior, deployment procedures) 5. **Existing rules** — read all `.claude/rules/*.md`. Identify files that mix domain + technical content 6. **Git log** — run `git log --oneline -50` to identify domain-related commits (keywords: fix, deploy, API, auth, config) ## Step 2: Classify findings For each piece of domain knowledge found, classify: - **Category**: deploy, api-integration, business-logic, data-model, auth-flow, external-service,