The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes system binaries including ffmpeg and google-chrome via child_process.execFile and child_process.spawn. These tools are required to encode video and perform headless rendering of the 3D icons.
[DYNAMIC_EXECUTION]: The loadPiece function in piece.mjs dynamically executes JavaScript code by copying .specimen.js files to a temporary directory and using the import() statement. This is the primary method used to evaluate the 3D scene definitions generated by the agent or provided by the user.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes and executes JavaScript files provided in the workspace.
Ingestion points: .specimen.js files are loaded via the loadPiece function in piece.mjs.
Boundary markers: The skill does not use specific boundary markers or "ignore instructions" warnings when processing these files.
Capability inventory: The skill possesses capabilities including file system writes (writeFileSync), arbitrary command execution (execFile, spawn), and dynamic code loading (import()).
Sanitization: Basic sanitization is performed via node --check to ensure syntax validity and a contract check to verify expected exports and data types.

specimen