The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided prompts and images which are sent to external AI models. This creates a surface for indirect prompt injection, although the risk is limited to the resulting image output.
Ingestion points: prompt and input_image parameters in scripts/generate_image.py.
Boundary markers: None; inputs are passed directly to the API request body.
Capability inventory: Performs network requests and file system operations (read/write) in scripts/generate_image.py.
Sanitization: No sanitization or filtering is applied to the input strings or image buffers.
[DATA_EXFILTRATION]: The skill communicates with the well-known AI service OpenRouter to perform its core functions.
Evidence: requests.post calls to https://openrouter.ai/api/v1/chat/completions in scripts/generate_image.py.
[DATA_EXFILTRATION]: The skill accesses local .env files to retrieve API credentials, which is a standard method for secret management.
Evidence: The check_env_file function in scripts/generate_image.py identifies and reads OPENROUTER_API_KEY from environment files.

generate-image