The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to process external, potentially untrusted data such as video transcripts and raw footage.
Ingestion points: Reads video files (raw.mp4) and processes text transcripts (Layer 2) through LLM instructions.
Boundary markers: No explicit delimiters or instructions to ignore embedded content are defined for the transcript processing step.
Capability inventory: The skill executes shell commands via ffmpeg and npx remotion, and performs network operations using the requests library in Python.
Sanitization: There is no explicit sanitization or validation of the content within processed transcripts before they are used to generate commands.
[DATA_EXPOSURE_&_EXFILTRATION]: The skill uses the ElevenLabs API but correctly manages credentials by retrieving the ELEVENLABS_API_KEY from environment variables rather than hardcoding it, which is a safe practice.
[EXTERNAL_DOWNLOADS]: The skill references and fetches assets from well-known services including ElevenLabs and fal.ai. These operations are legitimate and targeted at official API endpoints.

video-editing