The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines an interface that accepts React/TSX code through the code parameter. This creates an attack surface for indirect prompt injection, where malicious instructions embedded in external data (such as web pages or documents analyzed by the agent) could influence the generated video logic.
Ingestion points: The code and props parameters in SKILL.md serve as the entry points for untrusted content.
Boundary markers: No specific delimiters or instructions are present to prevent the agent from obeying instructions embedded within the provided code strings.
Capability inventory: The skill relies on the belt CLI to execute rendering tasks on a remote backend.
Sanitization: The skill does not specify any sanitization or validation routines for the input code before it is passed to the rendering tool.
[COMMAND_EXECUTION]: The skill uses the belt CLI to interact with the inference.sh platform. This involves executing shell commands that transmit the React component code for remote rendering.
[EXTERNAL_DOWNLOADS]: The skill references installation scripts and documentation hosted on the official inference-sh GitHub repository and domains. These resources are used to establish the local environment required for video rendering.

remotion-render