threejs-image-generator
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script (
scripts/generate_image.py) using theuvtool to perform image processing and file system operations. It reads input images and writes generated assets to user-specified paths within the project directory. - [INDIRECT_PROMPT_INJECTION]: The skill processes responses from Google's Gemini API. The Python script prints text responses from the model directly to standard output, which is then ingested into the agent's context. This creates a potential surface for indirect prompt injection if the external model is tricked into returning malicious instructions.
- Ingestion points: The
response.partsloop inscripts/generate_image.py(line 155) extracts text and image data from the API response. - Boundary markers: There are no explicit boundary markers or instructions to ignore embedded text when the model's text output is printed to the console.
- Capability inventory: The skill possesses the ability to read and write files on the local filesystem and perform network operations to communicate with the Gemini API.
- Sanitization: No sanitization or filtering is performed on the text returned by the external AI model before it is presented to the agent.
Audit Metadata