Infrastructure as code

Before adopting a service or changing infrastructure, find the repository’s declarative configuration and deployment path. Prefer a reviewable change to manual console work or one-off commands.

• Inspect the affected resource, dependencies, state, and existing conventions before editing. Keep credentials out of code, logs, and prompts.
• Preview the exact change with the repository’s supported tool. Read the proposed actions and errors; do not treat a successful preview as deployment.
• Apply only through the documented deployment path and verify the resulting resource or service behavior with observable evidence.
• If a change requires staged applies because later state depends on an earlier apply, keep those changes separate and explain the dependency. Never bypass declarative state with an ad hoc mutation.

Alert behavior

Treat a firing operational alert as an urgent signal. Confirm whether its target condition occurred; a false or stuck alert is a defect in the alert or the signal it reads, not a reason to silence it. Before deploying an alert change, verify that it fires for the target failure and stays quiet for healthy behavior. After deployment, check its state and promptly roll back or fix forward if it misfires. Provider and deployment adapters belong in user configuration or later skill bundles. This skill defines the workflow, not provider-specific commands.