ocr-and-documents

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime path can ingest outsider-authored free text when the operating user supplies a remote URL: web_extract(urls=[...]) fetches public web content (e.g., a PDF from arxiv.org or any arbitrary URL) and converts it to readable text/markdown that is then fed into the agent’s LLM context.