The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-controlled React/Remotion code through the code parameter and incorporates it into shell commands for execution via the belt CLI tool. This creates a surface for potential instruction injection or shell breakout if the agent does not properly escape the input.
Ingestion points: The code and props fields within the skill's input schema in SKILL.md accept untrusted data from the user.
Boundary markers: The instructions lack explicit delimiters or safety headers to separate user-provided code from the command structure.
Capability inventory: The skill utilizes the Bash tool to execute belt app run commands, which can be manipulated if the input is not sanitized.
Sanitization: No sanitization, validation, or escaping mechanisms are specified for the code input before it is interpolated into the bash command.
[EXTERNAL_DOWNLOADS]: The skill directs the agent to download and install additional components and configuration files from external sources.
Evidence: Instructions include npx skills add belt-sh/cli and point to installation scripts hosted on raw.githubusercontent.com/inference-sh. These resources are associated with the vendor's platform and are used for initialization.
[COMMAND_EXECUTION]: The skill's primary functionality relies on executing the belt CLI tool via shell commands to process rendering requests.

remotion-render