The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and process user-generated content such as 'Design briefs' and 'Core loop contracts' to build mechanics. This creates a surface where instructions embedded within design documents could potentially influence agent behavior.
Ingestion points: User-provided design artifacts described in SKILL.md.
Boundary markers: Templates provided for design briefs do not include explicit delimitations or warnings to ignore instructions within the data.
Capability inventory: The skill has the ability to write to the file system (via scripts/create_threejs_game.py) and execute shell commands for development (Vite/NPM/Playwright).
Sanitization: There is no explicit sanitization step for the design inputs before they are processed by the agent.
[COMMAND_EXECUTION]: The skill provides and uses a Python script (scripts/create_threejs_game.py) to scaffold new game projects, involving directory creation and file operations. Additionally, the instructions guide the agent to perform standard development tasks using shell commands, such as npm install, npm run dev, and playwright test.
[DYNAMIC_EXECUTION]: The testing utility scripts/inspect-threejs-canvas.mjs uses Playwright's page.evaluate to execute technical logic and metric collection within a controlled browser environment. It also uses dynamic imports to load dependencies like playwright and pngjs from the local environment.

threejs-gameplay-systems