The Agent Skills Directory

[SAFE]: The skill is well-documented and its functionality matches its stated purpose of managing image generation and editing tasks.
[EXTERNAL_DOWNLOADS]: The skill requires standard Python packages (openai, pillow) which are well-known and hosted on official registries. Documented usage of uv or pip for installation is standard practice.
[COMMAND_EXECUTION]: All operations are performed via the included scripts/image_gen.py. The script uses standard libraries to handle command-line arguments and interact with the OpenAI SDK.
[CREDENTIALS_UNSAFE]: The skill correctly handles sensitive credentials by instructing the user to set the OPENAI_API_KEY as an environment variable and explicitly warns against pasting the key into the chat.
[DATA_EXFILTRATION]: Network activity is restricted to official OpenAI API endpoints required for the skill's primary function. There is no evidence of unauthorized data collection or transmission to third-party servers.
[PROMPT_INJECTION]: The skill uses a 'prompt augmentation' technique that wraps user input into a structured template (e.g., 'Primary request: <user_input>'). This provides a layer of separation between user data and instructions.
[SAFE]: File system operations are localized to specific project directories (tmp/imagegen/ and output/imagegen/) for temporary and final artifacts.

imagegen