ClassifyCCP

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

• Third-party content exposure detected (high risk: 0.85). At runtime the skill ingests the user-provided PDF by extracting its page text via PyMuPDF into _tmp_converted.md, and that extracted prose is then read/translated/analyzed by the agent; if the PDF is authored by an outsider (e.g., a court judgment or authority document), its text becomes LLM context (indirect prompt injection risk).