---
description: Communication retrospective — honest analysis of where user and Claude failed each other this session
allowed-tools: []
---

Review the full session transcript above and produce a communication retrospective. Do not summarize what was built — only analyze how the collaboration worked or failed.

## Rules (non-negotiable)

1. **Symmetric:** You must produce the same number of observations for the user and for Claude. No padding either side.
2. **Evidence-bound:** Every observation must cite a specific moment — quote a message, reference a turn, name the exchange. No unevidenced claims.
3. **No generic advice:** Every recommendation must link to a specific incident from this session. "Communicate more clearly" is forbidden. "In the X exchange, you could have specified Y instead" is correct.
4. **No people-pleasing:** If Claude performed well and the user caused most friction, say so plainly. If Claude misread requirements repeatedly, say that. The report serves accuracy, not comfort.
5. **Length cap:** Under 400 words total. Cut ruthlessly.

---

## Report Format

### What Worked
One or two sentences only. Specific. Skip if nothing notable.

### User — Communication Gaps
- [observation tied to specific moment]
- [observation tied to specific moment]

### Claude — Execution Gaps
- [observation tied to specific moment]
- [observation tied to specific moment]

### Next Steps
**User should:** (bulleted, specific, incident-referenced)
**Claude should:** (bulleted, specific, incident-referenced)

---

Tone: direct, clinical, useful. No motivational language. No emojis. No closing summary.