food-diorama-skill
Pass
Audited by Gen Agent Trust Hub on Mar 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill does not exhibit any malicious behaviors such as data exfiltration, credential theft, or unauthorized persistence.
- [COMMAND_EXECUTION]: The script
scripts/generate_food_diorama.pyutilizes thesubprocess.runmethod to execute a secondary Python script (nanobanana.py) which acts as the image generation engine. This execution is performed using a list of arguments without a shell, which prevents command injection. The dependency is expected and resides within the user's local skill directory. - [PROMPT_INJECTION]: The skill ingests a city name as input and interpolates it into a prompt for the Gemini API. This input is validated against a local JSON database (
data/cities.json), which effectively sanitizes the input by ensuring only predefined city names are processed, thereby mitigating prompt injection risks. - [EXTERNAL_DOWNLOADS]: The README.md file contains standard installation instructions and references to external GitHub repositories for dependencies, including the author's own repository (
lqshow/claude-skills) and a plugin from a third-party developer. These are documented neutrally as required for functionality.
Audit Metadata