The Agent Skills Directory

[DATA_EXPOSURE] (SAFE): The script includes a load_env function that searches for .env files in standard locations (e.g., ~/.config/skills/.env) to load the OPENAI_API_KEY. This is standard practice for CLI tools and does not constitute unauthorized exposure.
[EXTERNAL_DOWNLOADS] (SAFE): The script uses urllib.request.urlretrieve to download generated videos. The source URL is obtained directly from the official OpenAI API response, and files are saved with a timestamped .mp4 extension.
[COMMAND_EXECUTION] (SAFE): No use of os.system, subprocess, eval, or other dangerous execution functions was found. The script uses structured API calls via urllib.
[INDIRECT_PROMPT_INJECTION] (LOW):
Ingestion points: The prompt argument is taken directly from user CLI input.
Boundary markers: The prompt is passed as a string in a JSON payload to the OpenAI API without specific delimiters.
Capability inventory: The script can perform network requests (urlopen) and write files to disk (urlretrieve).
Sanitization: No sanitization is performed on the prompt string. However, since the output is a video file saved locally, the risk of downstream execution is negligible.

video-generation