合同风险审查虾

合同风险审查虾 — 深度识别合同风险条款，给出修订建议，守住法律红线。 **当以下情况时使用此 Skill**： (1) 用户上传合同文件（PDF/Word/文本），要求审查风险 (2) 需要识别不平等条款、违约责任失衡、知识产权陷阱、管辖权不利等风险 (3) 需要对比我方模板与对方合同，找出被修改的关键条款 (...

Security Scan

The skill's requirements, files, and runtime instructions are consistent with a contract-risk-review tool and do not request unrelated secrets or perform unexpected actions.

Purpose & Capability

Name, description, and included reference libraries align with the stated purpose (identify risky clauses, produce revisions, compare with templates). No unrelated credentials, binaries, or config paths are requested.

Instruction Scope

SKILL.md limits behavior to parsing contract text, semantic clause analysis, risk classification, and producing revision suggestions. It explicitly refers to the included reference files for judgement. Note: it suggests OCR for scanned PDFs but does not specify an OCR tool or how OCR is performed — this is an implementation detail rather than an incoherence, but you should confirm how OCR/parse steps run in your environment before sending sensitive contracts.

Install Mechanism

Instruction-only skill with no install spec and no code files, so nothing is written to disk by the skill bundle itself. This is low risk and coherent for a documentation/analysis skill.

Credentials

The skill declares no required environment variables, credentials, or config paths. The analysis and outputs rely on local input files and included references only — proportional to the purpose.

Persistence & Privilege

always:false and normal agent invocation settings. The skill does not request persistent or elevated privileges and does not declare modifications to other skills or system-wide settings.

Assessment

This skill appears internally coherent for contract risk review, but before use: (1) confirm how PDFs/DOCX/scan OCR will be performed in your agent environment (the SKILL.md recommends OCR but provides no tool — ensure a trusted OCR is used); (2) remember uploaded contracts contain sensitive data — verify data handling/privacy policies of the host/agent and avoid sending highly confidential contracts to untrusted environments; (3) if you plan to enable cross-skill push notifications (the SKILL.md references other '虾' skills), confirm those target endpoints and permissions; (4) test on non-sensitive sample contracts to validate output format and accuracy; (5) for high-stakes legal decisions, treat outputs as draft suggestions and have a human lawyer review final changes.

核心理念：合同不是走形式，每个字都可能是坑，让 AI 先帮你踩一遍。

工作流程

1. 文本提取 — 解析 PDF/Word/文本，提取合同全文
2. 结构解析 — 识别标题、章节、条款编号、甲乙方信息、附件
3. 语义风险识别 — 逐条分析，识别 6 类风险（见下表）
4. 风险分级 — 🔴 高风险 / 🟠 中风险 / 🟡 低风险
5. 生成修订建议 — 每个风险点给出：风险说明 + 修改建议 + 替代文本
6. 输出审查报告 — 风险概览 + 风险清单 + 修订建议 + 对比版本

6 类风险类型

风险类型	识别要点	典型案例
不平等条款	权利义务明显失衡	"乙方无条件接受甲方单方面变更价格"
违约责任失衡	一方违约责任过重或过轻	"甲方违约赔偿1%，乙方违约赔偿30%"
知识产权陷阱	成果归属、使用权限不明	"项目成果归甲方所有，乙方不得使用"
管辖权不利	约定对方所在地法院管辖	"争议由甲方所在地法院管辖"（甲方在外地）
表述模糊	关键条款用词不明确	"尽快交付"、"合理费用"、"适当补偿"
隐藏义务	附件、从句中的额外义务	附件中要求"乙方承担所有税费"

输入要求

• 必填：合同文本（直接粘贴 / 上传文件）、合同类型
• 推荐：我方角色（甲方/乙方）— 影响风险判定方向
• 可选：我方模板（用于对比分析）

输出格式

## 合同审查报告

**总体评分**：A/B/C/D（A=低风险可签，D=高风险不建议签）
**风险概览**：高风险 X 处 / 中风险 X 处 / 低风险 X 处

---

### 🔴 高风险条款

**[条款位置]** 原文：...
- 风险说明：...
- 修改建议：...
- 替代文本：...

### 🟠 中风险条款
...

### 🟡 低风险条款
...

---
**审查结论**：...

关键设计原则

• 语义理解优先：不只是关键词匹配，理解条款的真实含义和逻辑陷阱
• 上下文关联：结合合同类型、我方角色判断风险方向
• 可解释性：每个风险点都有明确的理由和依据
• 实用性：给出可直接使用的替代文本，而非泛泛建议

参考资料

• 风险条款库：references/risk-clauses.md — 典型风险条款案例，按类型分类
• 合同模板库：references/contract-templates.md — 各类标准合同关键条款
• 法律法规库：references/legal-regulations.md — 相关法律条文要点
• 修订建议库：references/revision-suggestions.md — 标准修订建议和替代文本

遇到复杂合同或行业特定合同时，读取对应 references 文件以获取更精准的判断依据。

文件处理

• Word (.docx)：直接解析文本内容
• PDF：提取文本层；扫描件需 OCR（建议 300 DPI+）
• 纯文本：直接分析
• 批量审查：逐份分析，汇总高风险合同清单

协作关系

• 发现高风险条款 → 可联动跨渠道消息推送虾推送法务团队
• 审查通过后 → 可联动合规哨兵监控虾持续监控合同对方合规风险