The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external user-provided data in the form of video prompts and subsequently reads these prompts back from local JSON sidecar files using jq to facilitate video re-rendering or 'promotion'.
Ingestion points: User-provided prompts for video generation and metadata stored in .json sidecar files (SKILL.md).
Boundary markers: The instructions suggest the use of specific output delimiters such as === JOB i/N <id> === to maintain context separation.
Capability inventory: The skill performs local shell command execution via a vendor-supplied script to interact with APIs and local tools like ffmpeg (SKILL.md).
Sanitization: The skill employs jq for structured data extraction and strictly enforces a multi-step approval process where the user must review a cost table before any billable submission occurs.
[COMMAND_EXECUTION]: The skill relies on executing a local shell script (ofox-video.sh) located in a sibling directory or common installation path. It uses this script for model discovery, price estimation, and video generation tasks. The skill includes logic for an agent to probe multiple local filesystem paths to resolve the location of this dependency.
[EXTERNAL_DOWNLOADS]: The skill directs the agent and user to the vendor's official domain (ofox.ai) for API key management and uses standard package managers (npx) for dependency resolution. It explicitly instructs the agent not to perform silent installations, instead requiring user intervention for any missing system components.

shorts-reels