The Agent Skills Directory

[PROMPT_INJECTION]: The skill features a surface for indirect prompt injection as it interpolates user-controlled strings into the final prompts sent to the image generation model.
Ingestion points: Ingests user-provided text through command-line arguments and external prompt files read by scripts/image_gen.py.
Boundary markers: Employs labeled sections (e.g., 'Primary request:', 'Constraints:') to structure prompts, but lacks cryptographic delimiters or explicit instructions to the model to ignore potential directives embedded within the user input.
Capability inventory: The skill is capable of executing its own Python CLI, writing image and temporary data files to the disk, and performing network operations via the OpenAI API.
Sanitization: Input strings are largely unsanitized before being added to the API payload, with sanitization limited to filename-safe slugification for output files.
[EXTERNAL_DOWNLOADS]: The skill requires external dependencies and communicates with a well-known service provider.
Evidence: References and installs the openai and pillow Python packages from public registries.
Evidence: Makes legitimate outbound requests to OpenAI's official API domains to perform image generation and editing.
[COMMAND_EXECUTION]: The skill uses a bundled command-line utility to interface with the external API.
Evidence: Executes the internal script scripts/image_gen.py to handle the logic for image creation, editing, and batch processing.

imagegen