Record and apply per-seat reasoning settings, report read-only enforcement for the seats that will actually run, and attribute scorecard observations only to participating seats using their measured durations.
- Add a validated optional
reasoningfield to CLI roster seats. - Apply reasoning to Codex exec and app-server, OpenCode, Pi, and Grok.
- Preserve reasoning in roster snapshots and resume paths.
- Limit direct-worker read-only advisories to the selected worker.
- Base advisory classification on the effective transport and sandbox.
- Record per-seat execution duration in run artifacts.
- Count only actual worker and orchestrator participation in scorecards.
- Run focused tests and
./scripts/verifythrough Brigade.