peer-review

Runs cross-agent reviews of specs, plans, and code using Cursor, Codex, Claude, and Gemini CLIs.

1|Updated Mar 24, 2026
One-click install
npx skills add https://github.com/SgtPooki/skills --skill peer-review-sgtpooki
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: peer-review
Source: https://github.com/SgtPooki/skills/tree/main/skills/peer-review
Command: npx skills add https://github.com/SgtPooki/skills --skill peer-review-sgtpooki

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Getting a second opinion on specs, plans, code changes, or design decisions requires switching tools and manually copying context, and self-review by the same agent that wrote the work misses blind spots. This Skill dispatches review prompts to external CLI agents (Cursor, Codex, Claude, Gemini) so the reviewer is always a different agent than the author. ## Core Features & Use Cases - Multi-agent dispatch: Send a review prompt to one specific agent (cursor, codex, claude, gemini) or run all available agents in parallel and present each result under a labeled heading. - Context serialization: External CLIs cannot see the current conversation, so the Skill packages specs, issue bodies (via gh issue view), file paths, and diffs into the prompt, using temp files for long content. - Anti-recursion and grounding guardrails: Every prompt instructs the reviewer not to delegate to other CLIs while encouraging read-only file inspection, preventing recursive agent invocation and ungrounded reviews. - Use Case: After drafting an API design spec, run /peer-review codex What's wrong with this API design? to get Codex's critique, then post the feedback as a new comment while preserving the original spec for an auditable review trail. ## Quick Start Ask the agent to run /peer-review with a prompt like Review this spec for completeness to send it to all available reviewer agents in parallel.

Frequently Asked Questions about peer-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get a second AI agent to review my code or spec?▼

Run /peer-review followed by your review prompt to send it to all available agents in parallel, or prefix with cursor, codex, claude, or gemini to target one agent. The Skill serializes the needed context into the prompt since external CLIs cannot see the current conversation.

What is the difference between codex exec and codex review?▼

codex review --uncommitted is only for reviewing actual uncommitted changes in the current repo's working tree. codex exec handles everything else: spec review, plan review, design feedback, and arbitrary prompts with codebase context.

Which CLI tools does peer-review support?▼

It supports cursor-agent (Cursor), claude -p (Claude), codex exec and codex review (Codex), and agy -p (Gemini via the Antigravity CLI). Missing tools are skipped and reported; the run fails only if every tool is unavailable.

Why does the Cursor CLI not produce review output?▼

The binary must be cursor-agent, not cursor. The cursor command launches the GUI/Electron app, ignores flags like --print, and opens a window instead of printing a review. Also cd into the repo root first since cursor-agent has no --workspace flag.

How do I review a long spec or large diff with an external agent?▼

Write the prompt to a temp file first to avoid argv limits. Claude and Codex accept the file via stdin, while Cursor requires passing the file contents with command substitution since it has no stdin mode.

Why should the reviewing agent differ from the authoring agent?▼

Using a different agent avoids shared blind spots: if Claude wrote the work, review with Cursor or Codex, and vice versa. The Skill also injects an anti-delegation instruction so the reviewer produces its own review instead of recursively invoking other CLIs.