CedarCopilot wants to merge 1 commit into main from staging
Not deployed
Libra has no production signal for this change yet because it has not deployed. Libra checks hourly for 3 days after each deploy.
This change centralizes authorization, adds administered-user and document-owner support, expands agent and playbook tooling, and updates deployment planning to consider each surface’s last successful deployment. The cancelled-deployment and no-prior-success paths behaved as intended. However, a GitHub Actions history lookup failure can still make the planner omit frontend work from a cancelled deployment, leaving it undeployed until a later frontend change triggers another release.
Not safe to merge until deployment planning handles unavailable GitHub Actions history without skipping previously cancelled surface changes.
The affected release path was reproduced using a synthetic Git history and mocked GitHub Actions responses. The checks confirmed the intended successful-baseline and no-prior-success behavior while demonstrating the remaining lookup-outage failure.
Files Needing Attention: aws/scripts/plan-surface-scope.sh and aws/scripts/last-deployed-sha.sh
What T-Rex did
General comment
Baseline lookup failure can silently strand changes from a cancelled deployment
aws/scripts/plan-surface-scope.sh:95-97, an empty result from last-deployed-sha.sh uses the pushed baseline. last-deployed-sha.sh:63-66 intentionally returns empty on a GitHub API/list-runs failure. If a frontend deployment was cancelled at commit C, then a later unrelated push is planned with before=C, a lookup outage causes the planner to diff C..HEAD and report deploy_frontend=false; the frontend change at C remains undeployed until another frontend change occurs.github.event.before. Preserve the eLibra has not measured any production surfaces for this change yet.