The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill mentions the installation of the official openai Python library via pip. This is a trusted dependency required for the skill's functionality.
[COMMAND_EXECUTION]: The provided code examples demonstrate standard file system operations for reading input assets (such as room images and masks) and saving generated image data locally. They also initiate network requests to OpenAI's official API endpoints for image processing. These actions are within the expected scope of an image generation utility.
[PROMPT_INJECTION]: The skill functions as a gateway for user-provided instructions and image data to be processed by large-scale AI models, which inherently introduces a surface for indirect prompt injection.
Ingestion points: User-defined strings passed to the prompt parameter and image files/URLs passed to the image parameter in the client.images API calls within SKILL.md.
Boundary markers: The documentation provides structured prompting strategies to help distinguish instructions from the content being manipulated, such as explicitly labeling exact copy and using named indices for multiple inputs.
Capability inventory: The skill utilizes the OpenAI API to generate and edit images, combined with file system access for saving results and reading local assets (found in SKILL.md).
Sanitization: The documentation recommends using the API's built-in moderation="auto" setting and explicitly warns against attempting to evade safety blocks through obfuscation or reformulation.

openai-gpt-image