The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-supplied prompts and local configuration files to construct API requests.
Ingestion points: User-provided prompt argument and --image path/URL in scripts/generate_image.py; configuration settings from .imagencn.json in the current or home directory.
Boundary markers: The skill does not implement explicit boundary markers or safety instructions when interpolating the user prompt into the API request body.
Capability inventory: The skill has the capability to write files to the local filesystem (saving generated images) and perform network operations (communicating with AI provider APIs and downloading result images).
Sanitization: Output filenames are sanitized using regex in make_output_name, but the prompt content itself is passed directly to external APIs.
[EXTERNAL_DOWNLOADS]: The skill downloads generated image content from various AI platform endpoints, including Google, OpenAI, Alibaba Cloud, and ByteDance. These operations target well-known and reputable service providers.
[COMMAND_EXECUTION]: The skill instructions include a command to open a local documentation page (docs/models.html) using system utilities like open or xdg-open to provide the user with a model reference guide.

imagencn