菜单
Run four parallel read-only subagents that each review the same diff from a different lens — security, performance, correctness, and readability — then merge findings into one report. Use before merging large or risky PRs.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Generate or edit images using the OpenAI Image API (gpt-image-2). Use when the user asks to generate, create, draw, render, illustrate, mock up, or edit an image, icon, logo, mockup, illustration, OG image, blog hero, marketing asset, or similar visual. Also use when the user supplies a reference image and asks to modify, restyle, or remix it. Triggers on: "generate an image", "create an image", "make a picture of", "edit this image", "restyle this", "make a mockup of", "draw a", "render a", "illustration of".
When the same multi-step workflow repeats in Cursor (user corrections or agent redos), capture it as a new SKILL.md under .cursor/skills/ so future sessions load it automatically.
After navigating and interacting in Cursor's built-in browser, use browser_network_requests to audit every fetch/XHR for failures, slowness, duplicate calls, and suspicious payloads. Use for API-heavy pages and after backend or client networking changes.
When GitHub Actions fails, fetch failing job logs and assign each failing job to a separate subagent that fixes its slice of the problem in parallel. Use for multi-job CI failures where jobs are independent.
Execute a user flow step-by-step in Cursor's built-in browser while documenting each action, then emit a Playwright test that replays the same flow using stable selectors derived from the accessibility tree.
Automatically persist useful context — research, decisions, learnings, templates — to workspace files so knowledge survives across conversations.
| name | parallel-code-review |
| description | Run four parallel read-only subagents that each review the same diff from a different lens — security, performance, correctness, and readability — then merge findings into one report. Use before merging large or risky PRs. |
| user-invocable | true |
Use Cursor’s Task tool to run four explore (read-only) subagents at once. Each subagent only reads code and produces findings for one dimension. The main agent merges results into a single prioritized review — something only a multi-agent setup does efficiently.
Prefer a concrete list of files for reviewers:
git diff --name-only origin/main...HEAD
Or paste the PR link and let the main agent list changed files from the branch.
Send one message with four Task invocations, each subagent_type: "explore" and readonly: true, with prompts like:
Security
Read-only review: SECURITY
Changed files:
<list>
Focus: injection (SQL, shell, XSS), authZ/authN gaps, secrets in code, unsafe deserialization, path traversal, SSRF, IDOR, dependency CVEs mentioned in diff.
Output:
- Critical / High / Medium / Low findings
- File:line and short fix recommendation
- "No issues" if nothing material
Performance
Read-only review: PERFORMANCE
Changed files:
<list>
Focus: N+1 queries, missing indexes, accidental O(n²) loops, bundle size impact, unnecessary re-renders, sync I/O on hot paths, unbounded caches.
Output: same severity + location format as above.
Correctness
Read-only review: CORRECTNESS
Changed files:
<list>
Focus: logic bugs, off-by-one, wrong edge cases, race conditions, error handling gaps, breaking API changes, test gaps for new behavior.
Output: same format.
Readability / maintainability
Read-only review: READABILITY
Changed files:
<list>
Focus: naming, duplication, abstraction boundaries, file size, unclear control flow, missing types/docs where they would prevent bugs.
Output: same format; prefer suggestions over nitpicks.
The main agent should:
Fix in the main agent or spawn targeted non-readonly follow-up tasks only for approved items.