The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes workspace data including git logs, README files, and application source code to generate video scripts. This creates a surface for indirect prompt injection where malicious text in a processed repository could attempt to influence the agent's behavior. Ingestion points: git history and local project files. Boundary markers: Absent. Capability inventory: npm, npx, ffmpeg, python. Sanitization: Absent.
[REMOTE_CODE_EXECUTION]: The generate_voiceover.py script downloads audio content from the ElevenLabs API and subsequently processes the file using ffmpeg and ffprobe via subprocess calls. While this pattern was flagged as a potential risk, the source is a well-known service and the processed data is non-executable audio binary.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from established and reputable services, including synthesized audio from ElevenLabs and royalty-free music from Bensound and Pixabay. It also references framework-specific skills from the official remotion-dev organization.
[COMMAND_EXECUTION]: Extensive use of command-line tools to manage the video creation lifecycle, including npm and npx for project initialization and rendering, and ffmpeg for audio normalization, mixing, and final video assembly.
[DYNAMIC_EXECUTION]: The skill involves the generation and execution of code at runtime, specifically Python scripts for orchestrating voiceover production and React components for video compositions rendered by the Remotion engine.

promo-video