Managing Snowflake infrastructure manually leads to inconsistent access, untracked objects, and risky ad-hoc changes. This post presents a Cortex Code CLI-powered agent that wraps Terraform workflows in plain-English commands, letting engineers describe changes, review generated plans, and keep the final apply decision human-only. The agent includes skills for onboarding teams, detecting drift (unmanaged objects outside Terraform state), reviewing plans for Snowflake-specific destructive patterns, environment promotion, and safe teardown. Key safety features include interactive-terminal-only apply, name proposal before edits, protected resources, and enforced admin role separation. The full open-source repo is available on GitHub.
Sort: