Add mcp__rag-unified__search_memory + mcp__rag-unified__cross_project_search
vào tools list 4 agents (Investigator + Implementer + Reviewer + CICD Monitor).
Tại sao:
- Sub-agent spawn KHÔNG inherit MCP server access từ parent session
- 4 agents previously CHỈ có Read/Grep/Glob/Bash → re-read MD files manually
- Plan B pre-flight Investigator phải Read PE Mig 22-26 thủ công thay vì 1 RAG query
- Plan CA Reviewer Cat 1 wire claim verify KHÔNG retrieve historical gotcha cross-session
- Plan CA Hotfix 1 silent sidebar drop nếu Implementer có RAG → catch Pattern 16-bis trước commit
Trade-off accepted (anh chốt full 4 agents):
- Token cost spawn cao hơn (~5-10K extra per RAG query)
- Risk noise dilute focus → mitigate by skill-specific prompt focus
Pitfall #1 reinforced (S27 multi-agent setup):
- Session đang chạy KHÔNG hot-reload registry
- Anh restart Claude Code CLI để spawn S30+ pick up MCP RAG tools
- Plan B Chunk D Implementer đang chạy dùng config CŨ (no MCP) — KHÔNG affect
Verify post-restart (Anh):
- Spawn test Investigator → call mcp__rag-unified__search_memory thử
- Pass = MCP tools loaded; Fail = YAML syntax issue (fallback wildcard mcp__rag-unified__*)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Bro re-emphasize prompt setup với rule Thứ 9 mới:
"Các task bất kỳ, BẮT BUỘC phải phân việc đầy đủ cho các Sub Agent đúng
vai trò, vị trí đã được chỉ định"
Retrospective S22: em main solo 6/10 task lẽ ra delegate được (vi phạm):
- Plan D F2 toggle UI → Implementer Case 2 (cookie-cutter mirror UsersPage)
- Plan C task 4 reflection regression test → Implementer Case 3
- Plan C task 1-3 14 service test catch-up → Implementer Case 3
- S22+2 seed 20 user script → Implementer Case 1 (mass deterministic)
- S22+3 rename users SQL transaction → Implementer Case 1
- Plan F pre-flight prod sqlcmd → Investigator (read-only audit)
Em main solo ĐÚNG 4/10 task:
- Plan E phân quyền strict V2 (security cross-stack)
- S22+1 disable 3 button bug fix (reasoning chain)
- S22+4 attachment view + Section adjust (UX + schema decision)
- S22+5 Mig 30 per-NV (cross-stack schema refactor)
Forward S23+ rule enforcement:
1. .claude/agents/README.md +banner "🚨 RULE BẮT BUỘC (Thứ 9)" trên top
invocation tree + retrospective S22 lesson + workflow forward
2. docs/HANDOFF.md S22 chốt v2 Last updated với note S23+ forward
3. Plan B Contract V2 wire pre-allocated 5 chunk sub-agent role table:
- Pre-flight Contract V1 state → Investigator
- Chunk A Mig 31 schema → Implementer Case 2
- Chunk B Service ApproveV2Async → Em main Solo (cross-stack)
- Chunk C Mig 32 LevelOpinions + service hook → Implementer Case 2
- Chunk D FE ContractCreate Workspace V2 → Implementer Case 2
- Chunk E FE ContractDetail Section 5 V2 → Implementer Case 2
- Pre-commit each chunk → Reviewer
- Post-deploy → CICD Monitor
Em main solo CHỈ khi: schema/UX/architecture decision + cross-stack tight
coupling + bug fix reasoning chain.
KHÔNG cắt narrative cũ — append rule banner ở top + Plan B section trước
Session 21 timeline.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
