peer-review

# /peer-review - External Manuscript Review ## Purpose Produce a reviewer-style assessment of another author's work. This skill judges the manuscript as submitted, not as the author wishes it to be. Use `/argument-governance` first when the review needs a formal gap-contribution or claim-evidence map. ## Codex-Only Baseline Complete the review with Codex using the supplied manuscript, supplements, rubric, and source anchors. Do not require Gemini, gemini-agent, a second model, or a subagent. If another reviewer tool is available, record its output as optional advisory context only. If `/argument-governance` is unavailable in the user's environment, manually extract the same spine in the review report: problem, gap, contribution, main claims, evidence, boundary language, and reviewer risks. ## Enhanced Advisory Mode If the user explicitly provides or enables an API-key-backed model review, use it only after Codex has made a first-pass review from the supplied materials. Rules: - keep API keys in environment variables, never in review files - send only user-approved manuscript or supplement excerpts - label the external result as `advisory_external_review` - compare it against Codex's local findings - include only grounded, source-checkable points in the final review - do not treat the external model as a reviewer consensus, venue decision, or factual verifier ## Core Rules 1. Review only the manuscript and explicitly supplied supplementary materials. 2. Do not use ...