cursor-cli-review

What problem does it solve?

Code review and development planning using Cursor CLI in non-interactive mode. Three modes: (1) Standard Review — spawns 1-3 Cursor reviewers for correctness, design, and security; (2) Adversarial Review — runs a dedicated skeptical reviewer that pressure-tests the chosen approach and failure modes; (3) Plan — spawns 1-3 Cursor planners for architecture, implementation, and risk analysis. Default model: opus-4.6. ONLY trigger this skill when the user explicitly mentions "cursor" in their request — for example "use cursor review", "cursor adversarial review", "用cursor review", "让cursor帮我做计划". Do NOT trigger on generic "review my code" or "make a plan".

Core Features & Use Cases

• Spawn Cursor CLI agents in non-interactive mode to either review code changes or produce development plans, each from clearly separated perspectives.
• Implement three modes: Standard Review, Adversarial Review, Development Planning, with a default model.
• Enforce hard constraints: all analysis must run via Cursor CLI using agent -p; read-only review; gate on explicit trigger.
• Support deterministic target selection, structured JSON output, and safeguarding against default agent exposure.

Quick Start

Spawn Cursor CLI agents to review code changes or generate development plans in non-interactive mode.