agentleFS
adamos486adamos486/skills/review-packet/SKILL.md

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

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    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.

The file

Files
SKILL.md

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's requirements.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

  1. Locate the design. Newest docs/superpowers/specs/*-design.md; if there's no spec, reconstruct from this conversation.
  2. 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).
  3. Fill every REQUIRED slot in references/packet-template.md (including §0 routing and §9–§11).
  4. Write to docs/superpowers/specs/YYYY-MM-DD-<topic>-review-packet.md. Do not commit.
  5. 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)

MistakeFix
No current-status / next-actionA 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 preambleFed 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 conventionsThe new thread violates them (commits, skips the next skill). List no-commit + the next step
Rejected alternatives droppedThe reviewer can't attack reasoning they can't see — include considered/rejected + why
Echoing only the user's flagged doubtsProactively surface the subtle risks nobody raised
Linking instead of inliningInline it; an external reader has no access
Committing the fileLeave it for the user to commit

Reference

  • references/packet-template.md — the section skeleton with REQUIRED slots. Fill all of them.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.