bootstrap

# Bootstrap Feature Tree

Two-phase process to populate Feature Tree from an existing codebase:
- **Phase 1**: Feature Discovery (bottom-up, code → features)
- **Phase 2**: Workflow Identification (top-down, features → workflows)

Both phases are optional. Run Phase 1 alone, skip to Phase 2 with manual guidance, or run both.

---

## Phase 1: Feature Discovery

### Step 1.1: Module Detection

Scan the codebase structure and propose modules:

1. Read top-level directories (especially `src/`, `lib/`, `app/`)
2. Check config files (package.json, pyproject.toml) for hints
3. Present proposed modules to user:

```
I found these potential modules:

| Module | Path | Notes |
|--------|------|-------|
| auth | src/auth | login, session, oauth files |
| payments | src/payments | stripe, checkout |
| users | src/users | profile, settings |
| utils | src/utils | shared helpers → likely INFRA |

Add, remove, or rename any? [Enter to confirm]
```

Log module detection: `bootstrap_log("Proposed modules: auth, payments, users, utils", "MODULE_DETECT")`

### Step 1.2: Parallel Subagent Dispatch

For each confirmed module, spawn an Explore subagent with Feature Tree context:

```python
Task(
    subagent_type="Explore",
    prompt=f"""
## Context: Feature Tree Bootstrap

Feature Tree tracks atomic features and infrastructure for AI-assisted development.

**FEATURE** = Atomic, user-facing capability
- Something you'd say "implement the X feature"
- Examples: AUTH.login, PAYMENT.checkout, USER.profile_update
- Must be completable in one Claude session

**INFRASTRUCTURE** = Shared utilities, not user-facing alone
- Use INFRA.* prefix: INFRA.rate_limiter, INFRA.logger, INFRA.validation
- Gets linked to features via `uses` field

**CONFIDENCE:**
- HIGH: Obvious from names/structure (file literally named login.ts with handleLogin export)
- MEDIUM: Reasonable inference from patterns
- LOW: Ambiguous, might be wrong

## Your Task

Analyze module: {module_name}
Path: {module_path}

1. Identi