A
planned · by CNE Associates
Agent Autopsy
Six real ways agents fail in production (loops, cost blow-ups, tool misuse, stale retrieval, prompt injection, silent regressions), replayed from traces, with the fix for each.
What it measures
- The cost counter during each replayed failure
- The after-trace once the fix is in
- The eval or CI gate that would have caught it
Who it is for
Teams with a stuck pilot who want to know what goes wrong after launch, and whether they would notice.
Shown in the demo gallery as Agent Autopsy: Six real ways agents fail in production (loops, cost blow-ups, tool misuse, stale retrieval, prompt injection, silent regressions), replayed from traces, with the fix for each.