The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a Node.js shell command template used to calculate 3D model bounding boxes by reading and parsing local binary GLB files. It also instructs the agent to use npx sdk-commands build for project validation.
[DYNAMIC_EXECUTION]: Bounding box calculation is performed using a JavaScript snippet executed dynamically via node -e. This is a utility for the agent to analyze model extents and is not associated with remote or untrusted code execution.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing untrusted project data (such as scene.json and 3D model files) to influence entity placement and scene logic. It includes mandatory validation steps to mitigate risks associated with malformed or adversarial project data.
Ingestion points: The agent is instructed to read parcel data from scene.json and metadata from 3D assets in the assets/ directory (documented in SKILL.md).
Boundary markers: The skill does not define explicit delimiters for external data, but it enforces strict spatial validation against computed parcel bounds.
Capability inventory: The skill uses shell execution for model analysis and file-write operations to update scene configurations (documented in SKILL.md).
Sanitization: The logic includes robust checks for entity positions, scale, and parent hierarchies, as well as a final validation gate via the SDK build tool.

composites