CedarCopilot wants to merge 2 commits into staging from feat/playbook-instruction-eval-suite
Live on prod, no production signal yet
Verification cited only 10 matching post-deploy events, below the 20-event floor for calling a change verified. Nothing is failing; there is not yet enough traffic to confirm it.
Verification cited only 9 matching post-deploy events, below the 20-event floor for calling a change verified. Nothing is failing; there is not yet enough traffic to confirm it.
Strict CloudWatch fallback saw 1,782 success-shaped log lines matching slack, channel, channels, workspace, workspaces, but no tied operation was present, so Libra is not calling this working.
Libra has verdicts on 0 of 3 tracked behaviors on prod; 3 are still being checked. Libra checks hourly for 3 days after each deploy.
playbook-instruction-eval skill's eval suite: AOP classification, approval/task parity, deal-close task cancellation, domain-return-shape, notify gating, orchestrator/sequencer, and stage-channel-override evals, plus fixtures/baselines each depends on.eval-instructions.ts with the instructions for the new evals.docs/cedar-configuration-knowledge.md, the configuration reference these evals were authored against.This PR is not safe to merge because multiple advertised evaluators cannot run, several core measurements use non-production prompt contracts, and production-derived customer data is exposed.
### Issue 1
apps/server/.claude/skills/playbook-instruction-eval/eval-instructions.ts:83-100
This evaluator reads four meeting fixtures that are not tracked or generated by a setup step, so the documented command fails with `ENOENT` before any trial executes. The same omission affects `scout-pilot-stage-event.txt` in `stage-channel-override-eval.ts` and both sequencer th
Libra has not measured any production surfaces for this change yet.