deepseek-integration

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection attacks because it interpolates untrusted chat history directly into the LLM prompt.
  • Ingestion points: The context_messages parameter in the build_chat_messages function (SKILL.md).
  • Boundary markers: Absent. The prompt construction uses simple string formatting without delimiters or specific instructions for the AI to ignore instructions within the user-provided context.
  • Capability inventory: The skill performs HTTP POST requests to the DeepSeek API and emits events to the application frontend (SKILL.md).
  • Sanitization: Absent. There is no validation or escaping of the user content before it is processed by the AI model.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 12:25 AM