generate-lego-3D-build-plan

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

• Third-party content exposure detected (high risk: 0.70). The CLI and scripts (scripts/generate_lego_build_plan.py and .ts) accept arbitrary http/https reference image URLs which are sent in the payload to the external OneKey Router endpoint (https://agent.deepnlp.org/agent_router), meaning the remote agent is instructed to fetch/ingest untrusted third-party content that can influence generation.

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

• Potentially malicious external URL detected (high risk: 0.80). Both the Python and TypeScript scripts call the external runtime endpoint https://agent.deepnlp.org/agent_router to generate and return the LEGO build-plan (the service’s response directly provides the agent’s instructions/outputs), and the skill requires this external API at runtime.