hugging-face-paper-publisher

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

• Third-party content exposure detected (high risk: 0.75). The runtime link workflow downloads README.md from an outsider-controlled Hugging Face repo via hf_hub_download(...), then reads it as text (content = f.read()) and incorporates it into the updated README that is uploaded back—this is outsider-authored free text entering the agent’s LLM context if the agent later summarizes/uses content.