The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill translates user requests into executable Blender Python code and shell commands, creating a surface where malicious instructions in user data could potentially influence agent behavior.
Ingestion points: User queries such as "render this", "produce an image", or "save the result".
Boundary markers: None identified in the skill instructions to delimit user-provided content from the command templates.
Capability inventory: The skill utilizes mcp__blender__execute_blender_code for script execution and the Bash tool for file system operations and video processing.
Sanitization: No explicit escaping or validation of user-provided parameters is implemented before they are interpolated into the rendering recipes.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform file verification and post-processing tasks.
Evidence: Recipe 5 suggests using ls -la /tmp/output_hero.png to verify the existence of a rendered frame.
Evidence: Recipe 6 provides a template for using ffmpeg to encode a sequence of PNG frames into an MP4 video file.

blender-rendering