Every tool the agent can call, with its safety class and dry-run behaviour.
Commands that exist only in the CLI, with no API operation behind them.
Every POSTSALE_ variable, what it controls, and its default.
What each exit code means and how a script or an agent should react.
Known limits of the CLI and the platform, stated plainly.
The four safety classes and what each requires before it runs.
Every schema slug the CLI can describe, grouped by command.