Scenario catalog

A scenario is a reproducible fault, injected into a live use-case stack, paired with a closed-loop check that decides whether an agent actually restored the service. Scenario identities and the methodology around them are public; that is what this page is. So are the pruned run records committed beside each scenario, which carry the verdict, timings and score. What stays private is the authored root cause and the oracle rubric for each scenario, together with the full transcript-bearing record that each pruned record references by hash (locally, per-run detail is rendered by pnpm runs:dashboard). The Role column says what each scenario is for today: most no longer separate frontier models on capability, so they earn their keep as regression guards rather than as discrimination axes.

Scenario	Use-case	Profile	Role	Status
cascading-upstream-failure	booklogr	incident	regression-guard / CI-smoke	DISQUALIFIED
compound-latency-pool-and-sort	booklogr	incident	certification-substrate	QUALIFIED
db-pool-exhaustion-deploy	booklogr	incident	regression-guard / CI-smoke	DISQUALIFIED
decoy-deploy-control	booklogr	incident	certification-substrate	QUALIFIED
latency-cache-stampede	booklogr	incident	regression-guard / CI-smoke	DISQUALIFIED
red-herring-coalert	booklogr	incident	regression-guard / CI-smoke	DISQUALIFIED
worker-cpu-starvation	booklogr	incident	regression-guard / CI-smoke	DISQUALIFIED