agentic-site-building
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFE
Full Analysis
- [DYNAMIC_EXECUTION]: The skill provides instructions for the generation and previewing of web code. It explicitly requires a 'safety boundary' where the generated code is rendered in an isolated sandbox, ensuring that sensitive data such as authentication tokens, storage, and privileged network access remain outside the runtime of the generated artifacts.- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risks of processing natural-language requests for code modifications. It addresses this attack surface by mandating explicit execution contracts, structured patches instead of full rewrites, and deterministic verification layers to ensure that only intended changes occur.- [DATA_EXPOSURE]: The orchestration contract includes a directive to persist state transitions and failure categories without storing secrets or unnecessary prompt content, demonstrating adherence to data minimization and security best practices.
Audit Metadata