160 Fn 15 3e52ac73

# fn-1.5 Integration testing

## Description
End-to-end testing of memory system in all three contexts:
1. Manual `/flow-next:plan` - memory-scout runs in parallel
2. Manual `/flow-next:work` - memory-scout runs during re-anchor
3. Ralph mode - both capture (PostToolUse) and retrieval work

Test scenarios:
- Memory disabled: no scouts run, no capture happens
- Memory enabled but empty: scouts return nothing gracefully
- Memory enabled with entries: relevant entries retrieved
- NEEDS_WORK feedback captured correctly

## Acceptance
- [ ] `/flow-next:plan` spawns memory-scout when enabled
- [ ] `/flow-next:work` includes memory in re-anchor when enabled
- [ ] Ralph capture writes to pitfalls.md on NEEDS_WORK
- [ ] Memory-scout retrieves relevant entries in Ralph loop
- [ ] Disabled config prevents all memory operations
- [ ] Works in smoke test environment

## Done summary
- Verified memory disabled prevents all operations
- Verified memory enabled allows add/read/list/search
- Verified ralph-guard extract_feedback and is_learnable work
- Verified skills reference memory-scout with config gating

Why:
- Confirms integration works end-to-end
- Tests both enabled and disabled states

Verification:
- Manual integration tests passed
- flowctl validate --all passes
## Evidence
- Commits: cfd27b75cd349ecf0f0ae91bffbcb4a9d4d6f7b4
- Tests: integration tests: disabled behavior, enabled operations, ralph-guard capture, skill references
- PRs: