/feedback

Team Feedback Registry

> **Purpose:** Accumulated learnings from /team runs. Layer 2 agents read > relevant entries before starting work so they don't repeat past mistakes. > > **How it's used:** The orchestrator reads this file, extracts entries > relevant to the current action/tools, and inlines them into Layer 2 > agent prompts alongside toolkit entries. > > **Entry lifecycle:** > - New entries appended by Layer 4 principals after every /team run > - Status: `pending` → `applied` (when toolkit quality notes updated) > - Entries older than 90 days with status `applied` can be archived > - Maximum 50 active entries (oldest applied entries archived first) > > **Context window budget:** This file should stay under 200 lines. > When approaching the limit, archive old `applied` entries to > `team:feedback-archive.md`.

---

Entries

2026-03-19 | /team review | framework cleanup (pre-redesign)

• **Tool reviewed:** Manual parallel agents (identical reviewer clones)
• **Reviewer role:** N/A (human observation)
• **Finding:** Identical clone agents produce redundant findings — 5 reviewers find the same 3 bugs instead of 5 different categories of issues
• **Recommendation:** Use diverse specialist roles with distinct review focus areas
• **Status:** applied (this redesign)
• **Impact:** high — fundamental architecture change

---

Entry Template

### YYYY-MM-DD | /team <action> | <scope>

- **Tool reviewed:** [skill or command name]
- **Reviewer role:** [which Layer 3 specialist found this]
- **Finding:** [what was wrong — specific, with examples]
- **Recommendation:** [specific fix — not "improve" but "add X to section Y"]
- **Status:** pending | applied
- **Impact:** high | medium | low