The Agent Skills Directory

Data Grounding via External Services: The skill retrieves architecture guidelines and best practices from official Google Cloud documentation and a Google-hosted MCP server (developerknowledge.googleapis.com). These references are used to ensure the accuracy and safety of generated recommendations.
Instructional Guardrails for Code Execution: The skill includes explicit instructions for the AI agent to never execute generated scripts (such as Terraform or gcloud commands) autonomously. It mandates obtaining unambiguous user permission before any runtime validation or resource provisioning occurs, reducing the risk of unintended infrastructure changes.
Indirect Prompt Injection Surface: By design, the skill ingests user requirements to generate technical blueprints. While this creates a surface for potential instruction injection via user data, the skill mitigates this by enforcing a strict multi-phase approval process and including steps to identify and resolve contradictions in user input before proposing solutions.
Phase-Separated Design: The workflow enforces a strict separation between discovery and architecture phases, which helps prevent premature technical decisions and ensures that non-functional requirements like security and compliance are addressed before a solution is designed.

google-cloud-solution-architecture