This is the actual faultline engine running in your browser via WebAssembly - not an animation. Edit the agent's Python, break its tools, and get real verdicts.
Engine: loads on first run (~6 MB, one time) - the same code as pip install faultline
Scenario: this agent keeps a warehouse stocked at 250 units. True count: 240, so the correct order is 10. faultline will break get_inventory() three ways - a bent number, an empty response, a truncated payload - and judge what the agent does. Actions are stubbed: no real order can ever fire.
The agent (editable Python - this exact code runs)
Ready for your own agent? pip install faultlineno LLM judge - same input, same verdict, every run