The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions include various shell commands for FFmpeg to perform deterministic video editing tasks such as cutting, concatenating, and reframing. It also utilizes npx remotion for rendering React-based video compositions.
[EXTERNAL_DOWNLOADS]: The skill facilitates interaction with external services including ElevenLabs for audio generation and fal.ai for media assets. These references target well-known service providers and include best practices such as using environment variables for API authentication.
[INDIRECT_PROMPT_INJECTION]: The workflow relies on the agent processing potentially untrusted external data in the form of video transcripts to plan video structures and generate edit decision lists.
Ingestion points: Video transcripts and recordings are ingested for analysis in Layer 2 (SKILL.md).
Boundary markers: The provided example prompts do not include explicit delimiters or safety instructions to ignore embedded commands within the transcript data.
Capability inventory: The skill utilizes shell command execution (FFmpeg, Remotion) and network operations (requests), which could be targets for manipulation via injected instructions.
Sanitization: No explicit sanitization or validation of the ingested transcript content is performed before processing.

video-editing