标签：flaky-tests

Reads failed CI runs, re-runs flaky and infrastructure failures once, and explains real ones in one PR comment (Jev)

Classify every failing CI test as regression, flaky, environment, test-data or unknown. Rules first, Jev (TypeSafe) for the ambiguous ones. Never reruns, never masks.

Compare System One decision models (Laya & Jev) on the same evidence: one contract, per-engine confidence gates, every decision stored with the engine that made it.