review-packet
Skill10 starsChanged 7 months agoMITView on GitHub
Use when a design, spec, or brainstorm needs to leave this thread — to be reviewed by another agent, to feed adversarial-review, or to catch up / hand off the work to a brand-new thread or a different AI. Triggers: "review packet", "context dump", "second opinion", "paste into ChatGPT/Codex/Gemini", "hand this off", "catch up a new thread", "continue this in another agent", "I'm running out of context".
1 other project uses a version of this.
Adopted by
Most recently taken up by
- microsoft/cat-agent-skillstheir own version, first seen Nov 8, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- microsoft/cat-agent-skills★ 101same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository12 other files it gives its agents.
- autobuildSkill
- dadjokeSkill
- golang-cli-reviewSkill
- leanplanSkill
- production-readySkill
- song-lyricsSkill
- streaming-recommendationsSkill
- superbuildSkill
- superplanSkill
- supervalSkill
- teachSkill
- walkthroughSkill
The file
Files
Review Packet
Overview
Package the current work into one self-contained, routed document that something outside
this thread can use: an external reviewer, the adversarial-review critic, or a brand-new
agent thread continuing the work. One file — a routing preamble, a shared context core, and a
continuation layer.
Core principle: a reader can only use what they can see. A reviewer can't critique reasoning that isn't written down; a fresh thread can't resume work whose status, file locations, and next step aren't written down. The packet carries both the reasoning and the state.
When to Use
- Hand a design to an external reviewer (paste into ChatGPT/Codex/Gemini).
- Feed
adversarial-review— its context core (§2–§7) becomes that skill'srequirements.md; it injects its own plan-critic. - Catch up / hand off the work to a new thread or a different agent (you're out of context; someone else continues).
- Not for: reviewing a code diff (use
/code-review).
Workflow
- Locate the design. Newest
docs/superpowers/specs/*-design.md; if there's no spec, reconstruct from this conversation. - Mine the conversation for what the next reader cannot see — both the reasoning and the state:
- original request, verbatim; background + invariants; the decision log (chosen, why, and what was considered + rejected/deferred, and why); the soft spots (proactive);
- current status — what's done / in-flight / pending / blocked; current rev; last decision;
- artifact & file map — absolute paths to the spec, this packet, the skill/code, the branch;
- next actions & how to resume — the immediate next step, what to read first, which skill/command to run, and the working conventions to honor (e.g. never commit; the brainstorm→writing-skills/writing-plans flow; project rules).
- Fill every REQUIRED slot in
references/packet-template.md(including §0 routing and §9–§11). - Write to
docs/superpowers/specs/YYYY-MM-DD-<topic>-review-packet.md. Do not commit. - Hand off. Print the path and which sections each consumer reads (review / adversarial-review / catch-up). Offer to print inline.
What Makes the Packet Work (the contract)
- Self-contained — inline everything; it works even for a reader with no repo access.
- Routed — the §0 preamble sends each of the three consumers to its sections, so the review wrapper never hijacks a catch-up or adversarial-review reader.
- Expose the reasoning — rejected/deferred alternatives with their why.
- Surface soft spots honestly — a dedicated "scrutinize these" section.
- Continuation-complete — a fresh thread can resume from it alone: current status, an absolute file map, and the immediate next action are all present and concrete.
- Adversarial (for the review consumer) — the wrapper tells an external reviewer to find
problems; for
adversarial-review, feed §2–§7 and let it inject its own critic.
Common Mistakes (from baseline testing)
| Mistake | Fix |
| No current-status / next-action | A continuation thread does the wrong thing — baseline fresh threads "produced a review" instead of resuming. State status + the immediate next action |
| Relative paths or "see the spec" | A new thread/agent can't open them — use absolute paths in the file map |
| Review prompt with no routing preamble | Fed to adversarial-review it collides with that skill's own critic; a catch-up thread adopts the reviewer persona instead of continuing. Add §0 routing |
| Omitting working conventions | The new thread violates them (commits, skips the next skill). List no-commit + the next step |
| Rejected alternatives dropped | The reviewer can't attack reasoning they can't see — include considered/rejected + why |
| Echoing only the user's flagged doubts | Proactively surface the subtle risks nobody raised |
| Linking instead of inlining | Inline it; an external reader has no access |
| Committing the file | Leave it for the user to commit |
Reference
references/packet-template.md— the section skeleton with REQUIRED slots. Fill all of them.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

