alibabacloud-sls-data-agent

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

• Third-party content exposure detected (high risk: 0.85). The runtime LLM context is fed by the user-provided --question (free-form text) into the CreateChat request body as messages[].contents[].value via build_message_value(question) → stream_chat() → /chat SSE, and this is outsider-authored input relative to the agent’s own system/bundled prompts.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill at runtime calls the SLS DataAgent OpenAPI (default host https://starops.cn-beijing.aliyuncs.com, overridable via SLS_DATA_AGENT_ENDPOINT) and consumes the SSE CreateChat stream whose returned messages directly determine the assistant's prompts/responses, making that external endpoint a required runtime dependency.