The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill specifies a dependency on openai==2.52.0 in its scripts. As the official package from the trusted vendor (OpenAI) has not reached this version, this requirement represents a potential supply chain risk, such as a dependency confusion attempt or the use of an unverifiable package source.
[METADATA_POISONING]: The skill's description and scripts reference a 'gpt-image-2' model and a related 'gpt-image' API. These do not correspond to publicly known products or services from the stated provider (OpenAI), suggesting deceptive or inaccurate metadata.
[COMMAND_EXECUTION]: The instructions direct the agent to execute local Python scripts via the uv tool to interact with external APIs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text or file content as prompts for an external image generation service.
Ingestion points: Data provided via the --prompt argument or read from files specified by --prompt-file in scripts/generate.py and scripts/edit.py.
Boundary markers: None. The skill interpolates the user-provided prompt directly into the API request without delimiters or safety headers.
Capability inventory: The skill has the ability to perform network requests to an external API and write generated files to the local file system.
Sanitization: Implements a 32,000-character length limit on prompts and trims surrounding whitespace from inputs.

gpt-image