npxrigorrun# start the runner and open the interface
Setting a project up — connecting a system and teaching a job — happens in the interface. Running,
gating and comparing are also available from the command line, which is the half CI needs.
Every project on this machine, including broken ones.
rigorrun run --project <id>
Run the project’s suite against its agent.
rigorrun gate --project <id>
The same, with a non-zero exit if the bar is missed. --agent <name> is needed only when the project has more than one agent; with several and none named, the gate refuses rather than picking one.
run · gate--project <id> --case <case-id>
Only the named cases, in the suite’s order; repeat --case for more. An id the suite does not have stops the run before it starts. The result says which cases it covered, and such a run never becomes the baseline. gate over some of the cases is never a release PASS: it exits 3 unless one of them fails.
Connect an agent that answers on an address and does each case’s work itself. Probed before it counts as connected. --allow-host names a remote host (https only); --header Name=secret takes a secret’s name, never its value; --body-template, --completion, --claim-path, --settle.
rigorrun agent list --project <id>
The project’s agents, and whether each one answers.
A pack is a system RigorRun ships its own client for. Its setup commands are its own, named by the
pack’s id as the first word — rigorrun <pack> --help lists them, and rigorrun --help ends with
the packs in your build.
A real recorded run, replayed offline: what the model said beside what the system held. --live runs the bundled example end to end now; --report <file> writes every case to a page.