← Back to Skills Marketplace
vlarkspur

师爷 - 三模型交叉评审

by vlarkspur · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
41
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install shiye
Description
师爷 — 三模型交叉评审引擎。用 Kimi/GLM/Qwen 三个大模型同时评审文档、代码、简历等,输出评分对比、扣分详情、三方共识和改进建议。
README (SKILL.md)

师爷 — 三模型交叉评审引擎

三个 AI 一起帮你检查内容,专挑你看不见的毛病。

安装

pip install --user openpyxl  # 可选,仅 Excel 评审需要

配置 API Key:

export SHIYE_API_KEY="your-deepseek-api-key"
export SHIYE_API_BASE="https://api.deepseek.com/v1"

或在 shiye.py 第 14 行直接填入。

用法

# 自动推断评审标准(推荐)
python3 shiye.py --sample ./my-document.md

# 指定评审标准
python3 shiye.py --criteria ./criteria/code-review.md --sample ./my-code.py

# 列出已有的评审标准
python3 shiye.py --list

评审流程

  1. 自动分析:AI 识别文件类型,自动生成 5-8 条评审标准
  2. 三评委打分:Kimi、GLM、Qwen 三个模型分别逐条打分(1-5分)
  3. 对比表格:并排展示三个评委的评分 + 平均分
  4. 扣分详情:红色=需改进(\x3C4分),绿色=已达标
  5. 三方共识:所有评委一致认为需改进的项目
  6. 分歧度分析:高度一致 / 轻度分歧 / 显著分歧

适用场景

  • 📄 文档质量评审
  • 💻 代码审查
  • 📝 简历评估
  • 📧 邮件/通知审阅
  • 📊 PPT 内容检查

示例输出

══════════════════════════════════════════════════════════════════════
  📊 三评委评分对比
══════════════════════════════════════════════════════════════════════

标准             Kimi     GLM      Qwen     平均
────────────────────────────────────────────────────────────
代码可读性        4        3        3       3.3
错误处理          5        4        4       4.3
函数单一职责      2        3        2       2.3
────────────────────────────────────────────────────────────
综合              3.7      3.3      3.0      3.3

配置

环境变量 说明 默认值
SHIYE_API_KEY DeepSeek API Key
SHIYE_API_BASE API 地址 https://api.deepseek.com/v1

注意事项

  • 需要 DeepSeek API Key(支持 Kimi/GLM/Qwen 模型)
  • 评审 16000 字以内内容,超长会自动截断
  • 三个评委可能给不同分数,分歧度越高越需要人工复核
Usage Guidance
Install only if you are comfortable sending the reviewed content to the configured API provider. Do not use it on secrets, proprietary code, resumes, regulated data, or confidential documents unless that external processing is approved. Prefer environment variables for the API key rather than editing the script, and provide an explicit criteria file when reviewing sensitive material so the tool does not auto-send a sample for criteria generation.
Capability Tags
requires-sensitive-credentials
Capability Assessment
Purpose & Capability
The core behavior matches the stated review purpose, but the artifacts claim Kimi/GLM/Qwen cross-review while all model calls are routed through a single configurable DeepSeek-style API base, which makes provider/data-routing transparency unclear.
Instruction Scope
The README uses an open-ended trigger and says it can review any text, with no exclusions for secrets, private code, personal data, regulated data, or system prompts.
Install Mechanism
There is no automatic install script and metadata declares no install steps; setup asks users to provide an API key by environment variable or directly in the script, which is functional but encourages a less safe credential handling option.
Credentials
The script reads a user-selected sample file or inline sample and sends up to 1000 characters for automatic criteria generation and up to 16000 characters for review to the remote API; that is purpose-aligned but under-disclosed for potentially sensitive inputs.
Persistence & Privilege
No persistence, privilege escalation, background worker, broad local indexing, destructive behavior, or hidden credential harvesting was found; it only creates a temporary criteria file for auto-generated criteria and attempts to delete it afterward.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install shiye
  3. After installation, invoke the skill by name or use /shiye
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
首次发布:Kimi/GLM/Qwen 三模型交叉评审引擎,支持自动推断评审标准、三评委打分对比、扣分详情、三方共识分析
Metadata
Slug shiye
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is 师爷 - 三模型交叉评审?

师爷 — 三模型交叉评审引擎。用 Kimi/GLM/Qwen 三个大模型同时评审文档、代码、简历等,输出评分对比、扣分详情、三方共识和改进建议。 It is an AI Agent Skill for Claude Code / OpenClaw, with 41 downloads so far.

How do I install 师爷 - 三模型交叉评审?

Run "/install shiye" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is 师爷 - 三模型交叉评审 free?

Yes, 师爷 - 三模型交叉评审 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does 师爷 - 三模型交叉评审 support?

师爷 - 三模型交叉评审 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created 师爷 - 三模型交叉评审?

It is built and maintained by vlarkspur (@vlarkspur); the current version is v1.0.0.

💬 Comments