catleop/llm-basic/code-review-triage
Claude Code and agent skill by catleop/llm-basic
- community
Checking that the SKILL.md file is still in the repository.
对 GitHub PR 进行分类,判断是否可以自动 Review 还是需要人工介入。 内部 skill,由 code-review 调用。也可单独触发: "classify this PR", "should this PR be auto-reviewed", "triage PR #N", "判断这个 PR 是否需要人工 review", "PR 分类", "评估 PR 复杂度"。 Make sure to use this skill whenever the user asks to classify, triage, or assess whether a PR needs human review.
Install
npx vskill@latest install catleop/llm-basic/code-review-triageSKILL.md
Code Review Triage
对 PR 进行快速分类,决定走自动 Review 路径还是需要人工介入。
输入
owner/repo:仓库标识pr_number:PR 编号- Memory 规则摘要(可选,由 code-review-memory 提供)
执行流程
Step 1:获取 PR 基本信息(0 LLM 调用)
并发调用:
mcp__github__get_pull_request:获取标题、描述、状态mcp__github__get_pull_request_files:获取变更文件列表
Step 2:硬规则检查(0 LLM 调用)
以下任一条件满足 → 立即返回 human_required,跳过 LLM 分类:
| 规则 | 条件 | 原因 | |
|---|---|---|---|
| 超大 PR | 变更行数 > 500 | 自动 review 效果差 | |
| Breaking Change | 标题/描述含 "breaking change"、"BREAKING" | 影响范围大,需人工评估 | |
| RFC/设计文档 | 标题含 "RFC"、"[RFC]"、"design doc" | 需要架构讨论 | |
| 安全核心文件 | 涉及 3+ 个安全核心文件 | 安全变更需人工确认 | |
| Draft PR | PR 状态为 draft | 尚未完成 |
安全核心文件判断:按文件计数,每个匹配以下任意模式的文件算1个(不是按模式组计数):
*/auth/*、*authentication*、*authorization**/security/*、*permission*、*access_control**secret*、*credential*、*password*、*token*(排除路径含test或mock的文件)
例:auth/login.py(1个)+ security/permissions.py(1个)+ utils/token_helper.py(1个)= 3个 → 触发 human_required
Step 3:LLM 分类(1 次 LLM 调用)
若硬规则未触发,调用 LLM 进行语义分类。
Prompt 结构:
置信度处理:
- confidence < 0.6 → 强制设置
review_type = "human_required",skip_reason = "置信度不足,建议人工确认"
Step 4:输出 ReviewPlan
输出格式
向用户展示分类结果:
读取 references/triage-rules.md 了解详细的分类规则和边界情况。
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingLlmLearning/code_review_agent/skills/code-review-triage/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/catleop/llm-basic/code-review-triage)