Skip to main content
Start at the review header:
  • queued, running, completed, failed, and cannot_run are durable states. A cannot_run receipt is a result, not a silent downgrade.
  • The requested tier can differ from the effective tier when workspace policy imposes a cap.
  • Correlation and review IDs connect the app, API, worker logs, and receipts.
Then inspect each finding. Confirm the file and line anchor, severity, confidence, evidence, and whether execution reproduced the behavior. A finding can be new, unchanged, remediated, or regressed against its repository baseline. Finally read the coverage panel and safe-to-merge explanation. Missing build capability, incomplete checkout, analyzer failure, or an untested behavior expands the blind region and can force abstention.
Dismissal and suppression are auditable workspace actions. Explain why a finding is not actionable so future baselines and precision analysis preserve the decision.
Last modified on July 28, 2026