HAREL ASAF / FAILURE LAB / v1.0.0

The agent says "done".
What actually happened?

Pick a failure, run the same scenario before and after a fix, and inspect the trace. No model, API key or credit card needed.

Synthetic simulation, not a real incident Local only · no input collection

Ready. Change the size or run with the default of 4.

A benchmark you can reproduce

24 paired fixtures: 3 scenarios × 8 sizes. Each pair compares the same state machine with and without a guard. Two baseline fixtures stay safe even without the fix. This is not a model ranking, accuracy measurement or savings forecast.

Engine sourceMethodology

No network, timing measurement or API cost.

unzip failure-lab-kit.zip
cd failure-lab-kit
node --test benchmark.test.mjs
node run.mjs > results.json