The Agent Skills Directory

[SAFE]: Secrets Management: The skill implements robust handling of sensitive credentials by instructing users to store API keys in a gitignored .env.local file. It also includes a specific 'secrets check' step that scans build output to ensure no keys or environment variables are accidentally leaked during deployment.
[SAFE]: User Consent Gating: The instructions enforce strict boundaries on autonomous behavior, explicitly requiring user approval in the chat before spending credits (e.g., ElevenLabs audio generation) or publishing the game to public hosting services.
[SAFE]: Multi-Agent Security Review: The skill utilizes a specialized team of agent roles (Critic, Genre Director, Narrative Director) to review all generated content, including design documents, code, and scripts. This internal review process acts as a validation layer for both quality and security before content is exposed to the user.
[SAFE]: Financial Safeguards: The audio generation tool includes mandatory safety features such as dry-run estimates, project-wide credit ceilings, and a request-hashing cache to prevent redundant or excessive spending on paid APIs.
[SAFE]: Trusted Integrations: All external services and deployment targets mentioned (e.g., Hostinger, Vercel, Netlify, GitHub) are well-known, reputable technology providers. The skill uses official connectors and standard CLI tools to interact with these services, avoiding unconventional or risky communication methods.
[SAFE]: Indirect Prompt Injection Mitigation: While the skill processes untrusted user input (game ideas), it employs designated reviewer agents to verify that the resulting game design and code adhere to the defined story rules and safety principles, effectively managing the attack surface.

build-a-game