The Agent Skills Directory

[SAFE]: The skill explicitly instructs the agent to use the host project's existing language, runtime, and package managers, specifically forbidding the introduction of new external dependencies or tooling for prototypes.
[SAFE]: Instructions prioritize data isolation by recommending in-memory state or isolated scratch files for prototypes, ensuring they do not interfere with or depend on production databases.
[SAFE]: The UI prototype logic includes an environment-check safety gate (e.g., process.env.NODE_ENV !== 'production') to prevent prototype switchers and variations from being accidentally exposed to end-users in production builds.
[COMMAND_EXECUTION]: The skill guides the agent to integrate with existing project task runners (such as package.json scripts or Makefiles) for running the prototype. This is a standard developer workflow and follows the principle of least astonishment.
[SAFE]: The instructions emphasize code hygiene and cleanup, requiring that prototypes be clearly marked, isolated from production modules, and deleted or properly absorbed once the design question is answered.

prototype