debrief
Debrief Mode
The user is returning cold. They've delegated cognitive labor to you. Make it effortless.
Your Job
Mine the conversation, synthesize status, present options. The user should scan, pick, and go — not re-read or think hard about what's next.
On Activation
1. Mine the conversation for:
- What was accomplished (concrete wins, completed work)
- What's in progress (started but not finished)
- What's blocked (and why)
- Decisions that were made (so they don't re-litigate)
- Open questions or pending items
- Implicit commitments ("we should also...", "don't forget...")
2. Synthesize into scannable status:
Use tables and bullets. No walls of text.
## Status
| Category | Item | Notes |
|----------|------|-------|
| Done | Feature X | Merged to main |
| Done | Fixed bug Y | Tests passing |
| In Progress | API refactor | 60% complete, paused mid-file |
| Blocked | Deploy | Waiting on credentials |
| Pending | Update docs | Not started |
**Key decisions made:**
- Chose approach A over B because [reason]
- Agreed to defer X until after launch
**Open questions:**
- Still need to decide on [thing]
3. Present clear next-step options:
Don't make them figure out what to do. Offer choices:
## What's Next?
1. **Continue API refactor** — pick up where we left off in `src/api.ts`
2. **Unblock deploy** — track down credentials issue
3. **Switch to docs** — lower priority but quick win
4. **Something else** — tell me what you need
4. Let them pick without typing:
Use AskUserQuestion with the options above. They tap a button, you execute.
Tone
- Concise. Scannable. Action-oriented.
- Assume zero memory. Don't reference "as we discussed" without restating.
- Lead with status, follow with options.
- Bullet points over paragraphs.
- Tables over lists when comparing items.
Anti-patterns
- Walls of text they have to parse
- Asking "what would you like to do?" without options
- Assuming they remember context
- Burying the status in narrative
- Making them type when a button would do
Enter debrief mode now. Mine this conversation, synthesize status, present options, and use AskUserQuestion to let them pick.
More from corygabrielsen/skills
mission-control
Coordinate complex multi-step work using task graphs and parallel background agents. Use when work requires decomposition, delegation, and long-running operations that may survive context compaction.
11loop-codex-review
Automated code review loop with progressive reasoning levels. Runs n parallel Codex reviews (configurable via -n), Claude addresses issues, climbs from low→xhigh reasoning until fixed point (all n clean). Human approval at each iteration.
11review-pr
Thorough, educational PR review process. Gathers context, explains changes, identifies issues systematically, and drafts constructive feedback collaboratively before submitting.
11orthogonalize-pr
Identify and separate orthogonal (independent) change sets within a decomposed PR. Creates branches for each set that could theoretically be reviewed or merged separately.
10decontextualize
Course-correct when artifacts overfit to conversation context. Invoke when examples, names, or language are too specific to current work.
10loop-review-skill-until-fixed-point
Iterate /review-skill on a target until fixed point. Runs review passes until all reviewers return NO ISSUES.
9