stanleyyoga123/incidentbench

IncidentBench is a Kubernetes incident-response benchmark: researchers run repeatable scenarios, observe a bundled or external response system, preserve run evidence, then grade and compare outcomes offline. The graph follows the Runner’s experiment flow through applications, traffic, fault injection, response, and archiving, then the Grader’s assessment and reporting. It also preserves the reference agents’ detect–investigate–remediate–verify–learn stages and the MCP evidence tools. Some subsystem links are documented in the README rather than demonstrated in sampled source; those edges cite that documentation.