The Agent Skills Directory

Standard Tool Integration: The skill leverages official gcloud and terraform commands for infrastructure diagnostics. These tools are used within the context of troubleshooting, and the skill includes explicit instructions to perform discovery in a strictly read-only manner, preventing unintended infrastructure changes.
Read-Only Constraints: A critical security control in this skill is the prohibition of mutating commands. The instructions explicitly forbid the use of terraform apply or any deployment-triggering gcloud commands, ensuring the agent remains in a dry-run and validation state.
Indirect Data Processing: The skill's primary function is to ingest and analyze deployment logs and Terraform HCL files. While these sources originate outside the skill's code and could potentially house indirect prompt injection attempts, the skill is designed to map these logs to specific component schemas and error categories, which limits the risk of the agent following arbitrary instructions found within logs.
User-in-the-Loop Remediation: When identifying fixes for IAM permissions or disabled APIs, the skill generates the appropriate gcloud commands as a response for the user to review. It does not attempt to execute these commands itself, maintaining clear boundaries for authorization.

infra-deployment-debugging