Dual PR Review

---

name: dual-review description: Run pr-review twice in parallel — once in this Claude session, once via OpenAI's Codex CLI — then merge the findings into unique-to-each / agreed-by-both buckets followed by both raw reports. Use when the user says "dual-review PR #N" or wants a second-opinion review. effort: max

You are running the pr-review skill twice — locally as Claude, and remotely via the codex CLI — and merging the results. Be terse. The user wants the union of findings, not two separate reports.

ARGUMENTS: <PR number> (with or without leading #).

Steps

1. Preflight. Run command -v codex via Bash. If it errors, stop and tell the user to install the OpenAI Codex CLI (npm i -g @openai/codex or equivalent) and authenticate.

2. Resolve PR number. Strip a leading # if present. Call it <N>.

3. Build the Codex prompt. Read .claude/skills/pr-review/SKILL.md and append a final line ARGUMENTS: #<N>. Write the combined text to /tmp/dual-review-prompt-<N>.md.

4. Launch Codex in the background. Use Bash with run_in_background: true. Pipe the prompt via stdin — the pr-review skill starts with --- (YAML frontmatter), which codex's arg parser treats as a flag separator and rejects with exit 2 if passed positionally:

   cat /tmp/dual-review-prompt-<N>.md | \
     codex exec --skip-git-repo-check --sandbox read-only --full-auto - \
     > /tmp/dual-review-codex-<N>.md 2>&1
   

   --full-auto is required for non-interactive background use (otherwise codex blocks on approval prompts). --sandbox read-only keeps codex from touching the working tree. You'll be notified when it finishes. Do not poll.

5. In parallel, run pr-review yourself by following .claude/skills/pr-review/SKILL.md for PR <N>. Produce the full report (template and all). Hold it in memory; do not print it yet.

6. When the background task completes, read /tmp/dual-review-codex-<N>.md. If it's empty or errored, skip to step 8 and note the dual leg failed — don't suppress your own findings to fit the template.

7. Merge. Walk both reports finding-by-finding. Two findings match if they cite the same file.go:line (±a few lines) or describe the same root cause in different words. Preserve each finding's original severity tag ([blocker]/[nit]/[question]) and original wording — do not paraphrase. For agreed-by-both findings, prefer the more specific phrasing of the two and cite both severities if they disagree (e.g. [blocker — Claude] / [nit — Codex]).

8. Print the merged report in this exact order:

   ## PR #<N> — Dual review
   
   ### Findings unique to Claude
   - **[blocker]** `file.go:42` — …
   
   ### Findings unique to Codex
   - **[blocker]** `file.go:99` — …
   
   ### Findings agreed by both
   - **[blocker]** `file.go:123` — …
   
   ### Description vs. diff
   <one line: union of both reports' findings>
   
   ### Recommendation
   <one short paragraph merging both recommendations>
   
   ---
   
   ### Raw — Claude
   
   <verbatim Claude pr-review output>
   
   ### Raw — Codex
   
   <verbatim Codex pr-review output>
   

   If a bucket is empty, write _None._ rather than omitting the header — the user wants to see at a glance whether each agent found something the other missed.

Don'ts

• Don't drop a finding because the other side missed it. Bucketing exists precisely to surface those.
• Don't paraphrase severity tags or finding text in the buckets — copy them verbatim. Only the trailing Description vs. diff and Recommendation are synthesized.
• Don't post the review (no gh pr review, no comments). Just print.
• Don't strip the raw reports — they're the audit trail for the merge.
• Don't run codex exec in the foreground; it blocks long enough that running pr-review yourself in parallel is the whole point.

Caveats

• Codex inherits cwd and shell env, so gh and git auth carry over. If gh auth status fails for Claude, it'll fail for Codex too.
• Codex output is whatever its model decides to produce; it may not follow the pr-review template exactly. Bucket what's there; don't try to rewrite it.
• If the two reports disagree on a fact (e.g. one claims a line panics, the other claims it's safe), put it in unique to <agent> rather than agreed-by-both. Don't arbitrate — that's the user's job.