The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves the agent processing external video transcripts to generate structured edit plans and executable commands. This creates a surface where malicious instructions embedded in the transcript text could influence the agent's logic or command generation.
Ingestion points: SKILL.md (Layer 2: Organization)
Boundary markers: Absent in the example prompt templates provided for transcript analysis.
Capability inventory: Shell command generation (FFmpeg, Remotion) and network operations (ElevenLabs API).
Sanitization: No specific sanitization or escaping mechanisms are described for the transcript ingestion phase.
[COMMAND_EXECUTION]: The skill leverages local command-line tools for core video processing tasks.
Evidence: Extensive use of ffmpeg for trimming, concatenating, and reframing video files, as well as npx remotion render for generating compositions.
[EXTERNAL_DOWNLOADS]: The skill facilitates asset generation through connections to well-known external media services.
Evidence: Integrates with ElevenLabs via https://api.elevenlabs.io for voice synthesis and utilizes fal.ai models for image and sound effect generation.

video-editing