The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill documentation encourages security best practices for credential management, specifically recommending the use of direnv and .gitignore to ensure API keys remain in the local environment and are not accidentally committed to version control.
[COMMAND_EXECUTION]: The image generation logic is encapsulated in a Node.js script that performs validation on output paths. The safeOutPath function ensures that generated image files are written only within the intended project directory, effectively preventing path traversal attacks.
[EXTERNAL_DOWNLOADS]: The skill utilizes the official @google/genai SDK to communicate with Google's generative AI services. This is a trusted dependency from a well-known service provider.
[INDIRECT_PROMPT_INJECTION]: The skill processes image descriptions from a local IMAGE_SPEC.md file. While this represents a data ingestion surface, the risk is mitigated by the fact that the prompts are passed to a specialized image generation model, and the skill provides an interactive mode for user review of each generated asset.
Ingestion points: IMAGE_SPEC.md (parsed in scripts/src/generate-images.js).
Boundary markers: Interactive review mode allows the user to inspect results before proceeding.
Capability inventory: File writing (PNG images) and network access to Google AI APIs.
Sanitization: File paths are validated via path.resolve and boundary checking against the project directory.

generate-images