The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from reference videos and their transcripts (processed via Whisper) to plan content structure and pacing. This creates a potential surface for indirect prompt injection if malicious instructions are embedded in the reference media. The skill manages this risk by requiring manual user approval for the generated script and scene frames before proceeding to video generation, and by instructing the agent to never invent facts or claims not present in the user-supplied brief.
[COMMAND_EXECUTION]: The workflow involves the execution of standard media processing utilities including ffmpeg for frame extraction and audio normalization, and whisper-cli for generating timing data. These commands are used locally for legitimate content processing tasks.
[DYNAMIC_EXECUTION]: The skill uses the Remotion framework to programmatically generate video. This involves the agent scaffolding React/TypeScript projects and rendering them based on structured templates. The process includes a verification step where the agent compares rendered output against the original plan to ensure consistency.

ai-content-generation