The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.Popen in scripts/render_stills.py to execute npx remotion still. This is a core functionality for rendering animation frames. The script implements input validation using regex ([a-z0-9-]+) for target and beat identifiers to prevent command injection.
[EXTERNAL_DOWNLOADS]: The workflow requires the installation of specific, version-pinned NPM packages (e.g., @remotion/transitions@4.0.522) and uses npx create-video for project scaffolding. It also pulls additional functionality from the remotion-dev organization.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data from storyboard.json to generate content for video scenes. This text is rendered into images which the agent is then instructed to inspect visually, creating a potential surface for indirect prompt injection. The skill includes step-by-step review instructions for the agent to verify layout and legibility, which serves as a manual checkpoint.
Ingestion points: The storyboard.json file is read by Python scripts and imported into React templates to populate scene text.
Boundary markers: None present for the beat.text fields used in scene templates.
Capability inventory: Includes file writing, package installation, and command execution for rendering.
Sanitization: Identifiers are checked against a strict whitelist regex; text content is interpolated into standard React templates.

vector-scene-layout