generate-lego-3D-build-plan
Fail
Audited by Gen Agent Trust Hub on Mar 30, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill's scripts contain a hardcoded API key used as a fallback if no environment variable is set.
- Evidence:
DEMO_KEY = "BETA_TEST_KEY_MARCH_2026"inscripts/generate_lego_build_plan.pyandscripts/generate_lego_build_plan.ts. - [DATA_EXFILTRATION]: The skill sends user-provided input (prompts and image URLs) to an external service for processing.
- Endpoint:
https://agent.deepnlp.org/agent_router(DeepNLP service). - Data transmitted: The user prompt, reference image URLs, and API authentication token.
- [COMMAND_EXECUTION]: The skill documentation and installation steps instruct the user to run shell commands to install packages and execute the agent CLI.
- Commands:
npm -g install @aiagenta2z/onekey-gateway,pip install ai-agent-marketplace, andonekey agent ...executions. - [EXTERNAL_DOWNLOADS]: The skill depends on external packages that are downloaded during the installation process.
- NPM Package:
@aiagenta2z/onekey-gateway(Scoped to the author's organization). - PyPI Package:
ai-agent-marketplace. - [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface as it processes untrusted user input and sends it to a remote API without sanitization or boundary markers.
- Ingestion points: The
--promptand--imagesarguments inscripts/generate_lego_build_plan.pyandscripts/generate_lego_build_plan.ts. - Boundary markers: Absent. User input is interpolated directly into the JSON payload.
- Capability inventory: Performs network POST requests to an external service.
- Sanitization: Only basic URL scheme validation (http/https) is performed on image links; no text sanitization for the prompt.
Recommendations
- AI detected serious security threats
Audit Metadata