mineru-ocr

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs users to provide their MinerU Token to the AI (e.g., "帮我配置 MinerU，Token 是：xxx" and "新的 Token 是：xxx"), which requires the agent to receive and likely embed that secret verbatim into config files/commands, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.80). This skill explicitly accepts and processes arbitrary remote document and webpage URLs (see SKILL.md) and the runtime script (.claude/skills/mineru-ocr/scripts/convert.js — functions convertRemoteUrlWithTokenApi / convertRemoteUrlWithLightApi and downloadRemoteImages) posts those URLs to external APIs, ingests the extracted Markdown, and even downloads remote assets, so it clearly fetches and acts on untrusted third-party web content.