The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied subjects, actions, and audience descriptions to optimize prompts for AI video models. This process creates a surface for potential indirect prompt injection. However, the skill implements a strong mitigation by requiring the agent to present a complete timed production script for user approval before any generation occurs.
Ingestion points: User creative intent and scene descriptions provided in the SKILL.MD context.
Boundary markers: Mandatory review and approval of the complete production script, dialogue, and clip boundaries before paid tool calls.
Capability inventory: Calls to the generate_video_clips tool for media creation.
Sanitization: Instructions to refine prompts, remove contradictions, and adhere to model-specific capability documentation.
[COMMAND_EXECUTION]: The skill uses specific tools such as opus_video_tools_whoami for identity verification and generate_video_clips for media generation. These operations are restricted to the functional purpose of the skill and utilize project-specific identifiers and status monitoring to ensure proper control flow.

video-director