The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a shell script scripts/download-stitch-asset.sh that executes curl to download assets from external URLs provided at runtime.
[EXTERNAL_DOWNLOADS]: The skill is designed to fetch design assets from external locations, including well-known services like Google Cloud Storage (storage.googleapis.com). References to well-known technology services are generally considered safe under standard guidelines.
[DYNAMIC_EXECUTION]: The skill generates Remotion video compositions at runtime by interpolating data from screens.json into React/Remotion component templates.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data sources like Stitch screen exports and HTML files, which could contain malicious instructions designed to influence agent behavior.
Ingestion points: The skill gathers data from .stitch/designs/ and exported Stitch web assets as defined in SKILL.md.
Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to disregard instructions potentially embedded within ingested design content.
Capability inventory: The skill has the capability to execute shell commands (via the download script) and generate executable React code for video rendering.
Sanitization: The provided code templates do not implement explicit sanitization or validation of the text fields (e.g., titles and descriptions) extracted from external design artifacts.

stitch-remotion