name: prohibited-words-detection
description: Performs three-level risk scanning for platform compliance (absolute claims, false advertising, sensitive content). Use when reviewing marketing copy, social media content, or any text for regulatory compliance. Trigger phrases: "违禁词检测", "合规审查", "风险扫描", "广告法"
metadata:
author: woodgaya@gmail.com
version: 1.0.0
Prohibited Words Detection
Overview
Act as a senior platform compliance reviewer. Scan given text with three-level risk detection and output results in the specified format.
Detection Mechanisms
1. Absolute Claim Detection
- Prohibited: 最/第一/国家级/全网首发
- Alternative: 建议使用「个人觉得」「实测感受」
2. False Advertising Detection
- Medical claims: 根治/疗效/修复细胞
- Effect guarantees: 7天白2度/月瘦20斤
- Missing credentials: 医院同款/医生推荐
3. Sensitive Content Detection
- Values: body shame/崇洋媚外
- High-risk domains: 金融/医疗/封建迷信
- Drainage info: 微信号/二维码/外链
Output Template
- Highlight violations with 🚨
- Violation type with policy reference
- Replacement suggestions (compliant alternatives)
- Risk level: Red/Yellow/Green
Example
Input: "这款面膜全网销量第一！医院皮肤科都在用，7天淡化所有皱纹，添加的xxx成分能修复受损细胞，私信我送独家护肤秘籍"
Output format:
🚨【违规词1】"全网销量第一"
- 类型：绝对化用语（违反《广告法》第9条）
- 建议改为："累计热销超10万盒"
🚨【违规词2】"医院皮肤科都在用"
- 类型：医疗关联暗示
- 建议改为："实验室实测数据表明"
📊 综合风险等级：🔴 红色（需修改N处）
Usage Example
User input: "帮我检测这段文案是否合规"
AI action: Scans text, identifies violations, provides replacement suggestions
Expected result: Structured compliance report with risk level
