The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard development commands to maintain quality and performance.
The agent is instructed to run npm test after implementing changes to ensure code integrity (SKILL.md).
The npm run inspect:canvas command is used to compare live diagnostics against the defined render budget (technical-art.md).
[SAFE]: The skill implements established industry practices for high-performance 3D graphics in the browser.
It provides detailed recipes for PBR materials, procedural geometry, and shader optimization using Three.js (shader-cookbook.md).
It establishes a clear technical art contract with performance targets for draw calls, triangles, and memory usage (technical-art.md).
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting external game design documentation.
Ingestion points: The skill requires reading an active Game Design Document (GDD) and approved target images to guide the polish process (SKILL.md).
Boundary markers: Explicit delimiters or warnings to ignore instructions within the ingested GDD are absent.
Capability inventory: The agent has the ability to execute shell commands (npm test), write results back to the GDD, and interact with the blender-mcp tool (SKILL.md).
Sanitization: There is no mention of sanitizing or validating the content of the GDD before processing.

threejs-polish-builder