story-short-scan

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly says the browser-cdp flow must extract a Bearer token from Chrome cookies and use it to call backend APIs, which requires the agent to read and handle a secret token directly.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). Contains deliberate extraction and use of an authentication token from the browser cookie (Admin-Token) and in-browser eval of arbitrary JS for scraping, which are sensitive actions that can be abused for credential theft or unauthorized API access if misused.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.72). 在 browser-cdp 采集模式下，LLM 会读取点众/黑岩采集到的页面或 API 返回文本（如书名、简介、标签等）并进入后续分析与报告生成，因此存在“第三方网站/第三方作者内容（outsider）被当作自由文本注入提示上下文”的间接风险。