The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts utilize standard system utilities such as curl for API interaction and jq for parsing response data. The generate-image.sh and generate-video.sh scripts include functionality to use the open command on macOS to display generated media files to the user.
[DATA_EXFILTRATION]: The skill implements an upload_file function that sends local files specified by the user to api.muapi.ai. This is the intended behavior for processing local assets in media generation workflows.
[DYNAMIC_EXECUTION]: Scripts use python3 -c to execute a small Python snippet for JSON encoding. This is used as a security measure to sanitize user-provided prompt strings before including them in the final JSON payload sent to the API.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted inputs through local files and remote URLs.
Ingestion points: Local files via --file/--image and remote URLs via --image-url in scripts such as create-music.sh and generate-image.sh.
Boundary markers: None; content is uploaded or passed as URL strings.
Capability inventory: Subprocess calls to curl, jq, python3, and open across all scripts.
Sanitization: All text-based prompt inputs are JSON-encoded using python3 before interpolation into request bodies.

muapi-media-generation