How to use this site

Every page is written for a coding agent, not a human. A human sends the agent a page URL and an outcome. The agent reads the page, collects the inputs, invokes the skill, and runs the verification.

Page contract

Block	What the agent does with it
Mental Model	What the page's topic is and why it exists.
Goal	The single outcome the page produces.
Prerequisites	What must be true before running. Agent checks first.
Skill	The exact skill to invoke, and which library ships it.
Inputs	Table of values. Source says human-provided or agent-derived.
Auth precheck	Before the Run body, resolve named targets for the page's auth surface (account | workspace | both). Run live CLI checks. Live Databricks and cloud account ids must match the human-named ids. Workspace surface also requires workspace host, workspace id, and workspace profile. On failure: tell the human it is blocked, give exact config or commands to fix, and stop. Do not invent SSO recovery mid-skill.
Run	The commands or skill invocation. When the page has a Naming step, present ≤3 naming options, wait for a pick, then map them to Terraform inputs before writing HCL.
Verify	A runnable check plus its expected output.
Where this fails	Silent-failure traps: symptom, cause, fix.
Next	Do next, optional manual fallback, and back to section.

Start at contents and pick the section whose "Pick if" matches the outcome.