start-3-1-1
Pass
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the Bash tool to stage scenario assets and install necessary Python dependencies from a requirements file. It also uses a custom CLI tool to record student progress.
- [EXTERNAL_DOWNLOADS]: The skill downloads established Python libraries, including the Google GenAI SDK and the Pillow image processing library, from the official Python Package Index (PyPI).
- [CREDENTIALS_UNSAFE]: The skill instructs the user to set up a Gemini API key in a local .env file. It includes specific safeguards for the AI agent to ensure the key is verified for existence without being exposed in the conversation logs or output.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external content, such as user-provided images and style descriptions from a local HTML library, which are then incorporated into prompts sent to the image generation model.
- Ingestion points: The skill reads local image files and a style library (style-library.html) to construct generation prompts.
- Boundary markers: The agent is guided by a specific teaching script that dictates how to process user input and scenario data.
- Capability inventory: The skill possesses file system access and the ability to execute shell commands via the Bash tool.
- Sanitization: The skill processes local data files as-is when constructing prompts for the model.
Audit Metadata