For AI agents
Safe automation and predictable command contracts.
Read before acting
Read AGENTS.md and docs/CLI-AGENTS.md in the selected source. Select the correct --home for the installation. Start with --help and read-only JSON diagnostics. Do not adopt a manually configured owner deployment or run a duplicate live runtime against its account/session.
lvct --help
lvct doctor --json
lvct health --json
lvct errors --since 1h --jsonExit codes and time budgets
0: command/check completed; 1: failed or degraded health; 2: invalid arguments; 130: Ctrl+C. Errors returns 0 after a successful read even when errors exist. Runtime log unavailability returns 1 while showing local CLI events. Keep long build sessions alive; logs -f, health --watch and SSH open are intentionally persistent.
Explicit actions
Do not send a broker order to test connectivity, regenerate an encryption key, delete volumes or blindly retry an ambiguous trade. Stop/restart and support uploads require user authorization. --yes confirms authorization already given; it does not create permission. Never print raw environment, Docker inspect secrets, sessions or broker credentials.