agentleFS
Sign inStart free
pushprove-maxpushprove-max/Commandeck/skills/maxreview/SKILL.md

maxreview

Skill3 starsChanged 6 days agoApache-2.0View on GitHub

Run an on-demand whole-branch Codex review outside a Builder unit. One adversarial pass, then review rounds until one finds no P0-P2 finding (at most 10), fixing and committing after every round. Use when the owner says "/maxreview" or asks for a max review of the current branch.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Stop when a round finds no P0, P1 or P2 finding, or after 10 rounds.

    line 34

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Whole-branch review

A whole-branch review that uses the same wording as the Builder review passes. The wording is never hand-written: wording.sh reads it live from scripts/builder-codex-pass.sh, so it cannot drift.

Wording

bash ${CLAUDE_PLUGIN_ROOT}/skills/maxreview/wording.sh adversarial   # the angles
bash ${CLAUDE_PLUGIN_ROOT}/skills/maxreview/wording.sh review        # the closing focus

Each prints its text, then the "list every finding you can defend" line, the standing checks and one severity line. If it exits non-zero (no pass script, or its wording markers or variables are gone), stop and tell the owner. There is no hand-written fallback.

Loop

  1. Pick <BASE>, the one base the whole run reviews against (the branch's merge base with main). Work only on committed code; commit before each round. Never push.
  2. Round 1 is the adversarial pass. Every later round is the review pass. Every round, not only the first, runs node codex-companion.mjs adversarial-review --base <BASE> "<focus>", with <focus> the output of the matching wording.sh call. The plain review command takes no focus text and would drop the Builders' wording. codex-companion.mjs is the newest ~/.claude/plugins/cache/openai-codex/codex/*/scripts/codex-companion.mjs.
  3. Proof bar at intake: a finding with no exact failing input or command is dropped, and does not hold convergence.
  4. Fix each kept finding, run the tests that cover it, and commit. Then start the next round.
  5. A fix that would change anything users see must stop and ask the owner, with a before and after. Do not apply it until the owner answers.
  6. Stop when a round finds no P0, P1 or P2 finding, or after 10 rounds. After round 10 with findings left, say so.

Fast lane

A Fast-lane worktree keeps using scripts/fast-lane.sh review; this skill does not replace it there.

Record

On convergence, write maxreview-r<N>-reports.md (<N> the last round) with each round's findings and what was done about each, ending with the single line VERDICT: GO. Without convergence, write no such line.

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.

Check your setup, free