Quick start
Run a first local incident path from reported pain through evidence, owner routing, and validation.
Before the first run
Confirm that the repository dependencies are installed, the playbook validator passes, and the Studio opens locally. Optional AI is not required for the deterministic simulator. Start with a known scenario so the team can learn the operating rhythm before using an unfamiliar incident.
npm install
npm run validate:playbooks
npm run typecheck
npm run test
npm run devStart with a precise pain statement
Describe what broke, who or what is affected, when it began, whether a change correlates, what proof already exists, and what must not happen. A strong intake is observable and bounded. "The platform is slow" is weak. "Interactive jobs on one queue exceed the normal launch time after the scheduler policy update; batch jobs on the peer queue remain normal" is usable.
Follow the evidence gates
Answer only what can be supported. Use unknown when the evidence is not available. Capture timestamps, healthy-peer comparisons, command results, and physical observations. Do not select the answer that merely sounds most likely.
Review the tunnel
The compact Live Fault Tunnel shows current progress. Expand it to see the full topology: reported pain, committed evidence, current gate, candidate boundaries, owner route, safety gate, guided move, and exit proof. Close the map or press Escape to return without losing state.
Produce the record
At the end of the run, review the resolution, command record, owner handoff, escalation packet, and exit criteria. Export a dossier when another team, vendor, or reviewer needs a stable evidence package.