---

name: autoreview

description: "Autoreview closeout: local dirty changes, PR branch vs main, parallel tests."

---

# Autoreview

Run Codex's built-in code review as a closeout check. This is code review (`codex review`), not Guardian `auto_review` approval routing.

Use when:

- user asks for Codex review / autoreview / second-model review

- after non-trivial code edits, before final/commit/ship

- reviewing a local branch or PR branch after fixes

## Contract

- Treat review output as advisory. Never blindly apply it.

- Verify every finding by reading the real code path and adjacent files.

- Read dependency docs/source/types when the finding depends on external behavior.

- Reject unrealistic edge cases, speculative risks, broad rewrites, and fixes that over-complicate the codebase.

- Prefer small fixes at the right ownership boundary; no refactor unless it clearly improves the bug class.

- Keep going until the selected review path returns no accepted/actionable findings.

- If a review-triggered fix changes code, rerun focused tests and rerun the review helper.

- Default to Codex review. If Codex is unavailable or exits with an error, the helper may fall back to `claude -p`; `pi -p` and `opencode run` are explicit reviewer/fallback options. The helper runs nested Codex review in yolo/full-access mode by default; use `--no-yolo` only when intentionally testing sandbox behavior.

- Stop as soon as the review command/helper exits 0 with no accepted/actionable findings. Do not run an extra direct `codex review` just to get a nicer "clean" line, a second opinion, or clearer closeout wording.

- Treat the helper's successful exit plus absence of actionable findings as the clean review result, even if the underlying Codex CLI output is terse.

- If rejecting a finding as intentional/not worth fixing, add a brief inline code comment only when it explains a real invariant or ownership decision that future reviewers should know.

- Do not push just to review. Push only when the user requested push/ship/PR update.

## Pick Target

Dirty local work:

```bash

codex review --uncommitted

```

Use this only when the patch is actually unstaged/staged/untracked in the

current checkout. For committed, pushed, or PR work, point Codex at the commit

or branch diff instead; do not force `--mode local` / `--uncommitted` just

because the helper docs mention dirty work first. A clean `--uncommitted` review

only proves there is no local patch.

Branch/PR work:

```bash

git fetch origin

codex review --base origin/main

```

Do not pass an inline prompt with `--base`; current CLI rejects `--base` + `[PROMPT]` even though help text is ambiguous. If custom instructions are needed, run the plain base review first, then do a local/manual follow-up pass.

If an open PR exists, use its actual base:

```bash

base=$(gh pr view --json baseRefName --jq .baseRefName)

codex review --base "origin/$base"

```

Committed single change:

```bash

codex review --commit HEAD

```

or with the helper:

```bash

.agents/skills/autoreview/scripts/autoreview --mode commit --commit HEAD

```

Use commit review for already-landed or already-pushed work on `main`. Reviewing

clean `main` against `origin/main` is usually an empty diff after push. For a

small stack, review each commit explicitly or review the branch before merging

with `--base`.

## Parallel Closeout

Format first if formatting can change line locations. Then it is OK to run tests and review in parallel:

```bash

.agents/skills/autoreview/scripts/autoreview --parallel-tests "<focused test command>"

```

Tradeoff: tests may force code changes that stale the review. If tests or review lead to code edits, rerun the affected tests and rerun review until no accepted/actionable findings remain. Once that rerun exits cleanly, stop; do not spend another long review cycle on redundant confirmation.

## Context Efficiency

Codex review is usually noisy. Default to a subagent filter when subagents are available. Ask it to run the review and return only:

- actionable findings it accepts

- findings it rejects, with one-line reason

- exact files/tests to rerun

Run inline only for tiny changes or when subagents are unavailable.

## Helper

Bundled helper:

```bash

.agents/skills/autoreview/scripts/autoreview --help

```

The helper:

- chooses dirty `--uncommitted` first

- otherwise uses current PR base if `gh pr view` works

- otherwise uses `origin/main` for non-main branches

- use `--mode commit --commit <ref>` for already-committed work, especially clean `main` after landing

- should be left in `--mode auto` or forced to `--mode branch` for PR/branch work; do not force `--mode local` after committing

- supports `--reviewer codex|claude|pi|opencode|auto`; `auto` runs Codex first

- supports `--fallback-reviewer claude|pi|opencode|none`; default is `claude`

- falls back only when Codex is unavailable or exits nonzero, not when Codex reports findings

- writes only to stdout unless `--output` or `AUTOREVIEW_OUTPUT` is set

- supports `--dry-run`, `--parallel-tests`, and commit refs

- runs nested review with `--dangerously-bypass-approvals-and-sandbox` by default

- keeps accepting `--full-access`; use `--no-yolo` or `AUTOREVIEW_YOLO=0` to opt out

- still accepts legacy `CODEX_REVIEW_*` env vars when the matching `AUTOREVIEW_*` var is unset

- prints `autoreview clean: no accepted/actionable findings reported` when the selected review command exits 0

## Final Report

Include:

- review command used

- tests/proof run

- findings accepted/rejected, briefly why

- the clean review result from the final helper/review run, or why a remaining finding was consciously rejected

Do not run another Codex review solely to improve the final report wording. If the final helper run exited 0 and produced no accepted/actionable findings, report that exact run as clean.