deepseek-chat

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes a subprocess wrapper that POSTs user-provided --prompt/--system text to DeepSeek’s https://api.deepseek.com/v1/chat/completions endpoint and returns the model output, meaning outsider-authored free text provided by the skill invoker is directly ingested by the LLM at runtime.