agentleFS
oborchersoborchers/proqi/.agents/skills/codex-review/SKILL.md

codex-review

Skill45 starsChanged 52 days agoMITView on GitHub

Run a bounded read-only implementation review with Codex as a native background subagent from Codex, or through the non-interactive Codex CLI from another harness. Never create a Herdr pane, tab, workspace, or agent for a reviewer.

91 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use addskill, apply-firmware and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • ryanportfolioSkill

    addskill

    Create, import, update, or install repo or personal skills: runtime ownership, resources, discovery checks.

    Also in 6 projects that use this

  • ryanportfolioSkill

    apply-firmware

    Bring current folder (plain folder, existing repo, old or partial firmware copy) up to full Harness Firmware in place. Mirror external repo → adopt-repo.

    Also in 6 projects that use this

  • +10 more from ryanportfolio

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 6 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 79 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run one independent Codex review without turning the reviewer into an interactive terminal participant.

    line 8
  2. Very specific:

    Treat uncommitted files as a moving snapshot.

    line 19
  3. Very specific:

    Never create or split a Herdr pane, tab, workspace, or worktree for it.

    line 26
  4. Distinctive:

    When the invoking harness is Codex, use exactly one native Codex subagent through the harness's subagent or collaboration mechanism.

    line 35
  5. Very specific:

    Do not run the review in the primary agent.

    line 37
  6. Distinctive:

    Only when the invoking harness is not Codex, launch one ephemeral, non-interactive Codex CLI process through that harness's managed process …

    line 43
  7. Distinctive:

    Use the exact requested model when one is given.

    line 45
  8. Distinctive:

    Wait for the native subagent or managed CLI process only when its result is needed.

    line 81
  9. Distinctive:

    Verify every finding against the current source.

    line 86

“Common” is measured: kept by at least 6 in 10 of the 79 edited versions of this file we could read. “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

Codex Review

Run one independent Codex review without turning the reviewer into an interactive terminal participant. The implementation owner remains the only interactive coding agent in its Herdr worktree.

Establish the review snapshot

  1. Resolve the repository, worktree, pull request, commit, diff, or files in scope. Read every applicable AGENTS.md and relevant product or architecture contract before composing the assignment.
  2. Record the absolute working directory, branch, exact HEAD, requested base, and git status --short. Preserve unrelated changes.
  3. Treat uncommitted files as a moving snapshot. Compare HEAD and status again after the review. A final stable-head verdict requires a fresh review if the reviewed files changed during the pass.

Reviewer execution is never interactive

A reviewer must run as a non-interactive background worker relative to Herdr. Never create or split a Herdr pane, tab, workspace, or worktree for it. Never use herdr agent start, herdr agent prompt, or another terminal multiplexer to host a reviewer. The reviewer inspects the implementation owner's existing worktree.

Determine the invoking harness before dispatch. A Codex agent hosted inside Herdr is still running in the Codex harness. Herdr is only the terminal host and does not change the harness identity.

When the invoking harness is Codex, use exactly one native Codex subagent through the harness's subagent or collaboration mechanism. This is mandatory. Do not run the review in the primary agent. Never launch the codex executable, including codex exec or codex exec review, from a Codex agent. If that Codex agent cannot spawn a native subagent, report the blocker to its parent session. The parent Codex session may dispatch the native reviewer into the same existing worktree. Do not fall back to the CLI, a Herdr reviewer, or primary-agent review.

Only when the invoking harness is not Codex, launch one ephemeral, non-interactive Codex CLI process through that harness's managed process runner. Use the exact requested model when one is given. Otherwise retain the configured Codex default. Use the review worktree as --cd, select the read-only sandbox, and capture complete standard output and standard error in one collision-free file under /private/tmp. A representative shape is:

codex exec --ephemeral --sandbox read-only --cd REPOSITORY_ROOT \
  --color never REVIEW_PROMPT > RESULT_FILE 2>&1

The process may continue after the runner initially yields, but it must remain owned by that managed process session until exit. Do not use shell &, nohup, an interactive Codex session, or a dedicated terminal or Herdr pane.

Give a strictly read-only assignment

The review prompt must include:

  • the exact scope, base, and recorded HEAD;
  • relevant contracts and acceptance criteria;
  • an instruction to inspect and report only;
  • a prohibition on editing, creating, deleting, renaming, formatting, or otherwise mutating tracked or untracked repository files;
  • a prohibition on commits, pushes, pull-request changes, comments, reviews, process control, and other external mutations;
  • severity-ordered findings with concrete file and line evidence;
  • a request to report no findings plainly when none are supported.

The reviewer may run read-only inspection commands and non-destructive tests when they materially validate a finding. Tests may create ordinary ignored build or temporary output, but the reviewer must not update snapshots, fixtures, lockfiles, generated sources, or repository state. It must never weaken a test or gate.

Verify the result

  1. Wait for the native subagent or managed CLI process only when its result is needed. Record its completion status. For a CLI review, read the complete result file even when the process exits nonzero.
  2. Re-record HEAD and git status --short and classify the result as stable or moving.
  3. Verify every finding against the current source. Treat it as a hypothesis, not authority. Classify it as confirmed, partially confirmed, rejected with evidence, already fixed during a moving review, or requiring a product decision.
  4. The reviewer never implements fixes. The implementation owner may fix confirmed findings when its assignment authorizes that work. A later pass must use a fresh subagent or fresh CLI process.

Report the reviewed scope, exact revisions, execution route, moving-snapshot status, process outcome, and verified findings. Keep a CLI result file until its findings are incorporated. Inspect its exact path before removing only that owned artifact.

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.