The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands to process media and interact with video generation platforms.
It employs ffmpeg and afconvert to extract and convert audio from video files, ensuring voice consistency across clips.
It uses the libtv CLI for model searching, node creation, and task management.
It provides instructions for integrating with other platforms like higgsfield and 即梦 (Jimeng) via their respective CLI tools.
[EXTERNAL_DOWNLOADS]: The skill manages the download of generated assets from remote platform servers.
It uses libtv download to fetch completed video segments and title cards into the local project directory.
It references external resources for model documentation and community director practices hosted on GitHub and official vendor domains.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes user-provided narration topics and scripts.
Ingestion points: Narration scripts, character descriptions, and visual style preferences provided by the user are incorporated into model prompts.
Boundary markers: The instructions mandate specific file structures (e.g., 旁白.md) and interpolation delimiters (e.g., {} for script segments) to separate user data from instructions.
Capability inventory: The skill has capabilities for file system operations, local command execution (FFmpeg), and network interactions via vendor CLI tools.
Sanitization: The workflow includes mandatory user checkpoints after script creation, character design, and the first clip's production, ensuring the user reviews outputs before batch generation proceeds.

animated-voiceover