The Agent Skills Directory

[Command Execution] (LOW): The skill's core functionality relies on the execution of a local Python script scripts/generate_images.py to automate image generation. While the script is part of the skill's intended purpose, local script execution requires trust in the skill's source.- [Indirect Prompt Injection] (LOW): The scripts/generate_images.py script reads markdown files from a user-specified directory and sends their content to an external AI model. This creates a vulnerability surface where malicious instructions embedded in those markdown files could influence the model's behavior. Evidence Chain: 1. Ingestion points: Reads files from the directory passed to --prompts-dir. 2. Boundary markers: None; the script extracts raw text from markdown files and passes it to the API. 3. Capability inventory: The script has file-write permissions to save images and network access to call the Gemini API. 4. Sanitization: Absent; the content is only stripped of specific headers but otherwise remains unvalidated.- [Credentials Unsafe] (SAFE): The script correctly uses os.environ.get('GEMINI_API_KEY') to retrieve sensitive credentials, avoiding hardcoded secrets within the codebase.

core-resources