alterlab-markitdown

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly fetches and ingests public, user-generated content (e.g., YouTube transcripts) as part of runtime conversion — see SKILL.md and references/file_formats.md with the example result = md.convert("https://www.youtube.com/watch?v=VIDEO_ID") — and those converted transcripts/text_content are then read and used by the provided scripts (e.g., scripts/convert_literature.py, scripts/batch_convert.py, scripts/convert_with_ai.py), meaning untrusted third-party content can flow into downstream processing and LLM-driven workflows.