The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external video sources via an API and presents the resulting metadata to the agent and user. A maliciously crafted video could potentially influence agent behavior through the returned title, hook, or virality reason.
Ingestion points: API response stored in RAW_RESULT in scripts/run-ai-clipping.sh and later parsed by jq.
Boundary markers: No specific delimiters or warnings are used when interpolating API-returned strings into the agent's output context.
Capability inventory: The skill possesses the ability to execute shell commands, perform network operations (uploads/downloads), and write files to the local system.
Sanitization: No sanitization or validation is applied to the metadata strings (titles, hooks, reasons) returned by the service before display.
[EXTERNAL_DOWNLOADS]: The skill relies on the muapi-cli package, which it instructs the user to install via npm. It also performs network requests to muapi.ai for video processing and file uploads.
[COMMAND_EXECUTION]: The skill executes shell commands using a wrapper script scripts/run-ai-clipping.sh. It invokes system tools including bash, jq, and the muapi CLI to perform its tasks.

muapi-ai-clipping