d6e-workspace-api-client

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

• Third-party content exposure detected (medium risk: 0.30). Interactive chat and execute-by-intent workflows ingest user-authored free text from the request body/session (e.g., POST /api/chat reads messages and assembles prompt context, and POST /api/workflows/execute-by-intent reads message) and then pass it into the instance LLM (streamText/generateText).

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

• Potentially malicious external URL detected (high risk: 0.90). The docs show server runtime endpoints that fetch/install external skill markdown (POST /api/workspaces/{id}/skills/install) and a public skill-markdown fetch (GET /api/v1/skills/{name}?workspaceId=) which are used to build agent system prompts at runtime, so remote content can directly control agent instructions.