Guide System
Why guided runs failed, from trace-analyzer's review of each run.
Why did guided runs fail?
Each failed guided run, judged against its guide: whose fault it was, why, and what the next guide should change.
629 guided runs failed in the rangeClick a group or a reason to list its traces.
The guide was wrong2.4%15 runs · +2.4 pts
The agent did not follow it73.8%464 runs · −26.2 pts
Right guide, followed, still failed23.7%149 runs · +23.7 pts
Could not tell0.2%1 run · flat
The guide was wrong
- A step in the route was wrong or missing46.7%7
- The guide's goal did not match the task33.3%5
- A popup or trap the guide did not warn about13.3%2
- A screen or control the guide names is wrong6.7%1
The agent did not follow it
- Another wrong action or early stop47.8%222
- Repeated actions that changed nothing31.5%146
- Tapped something other than it meant to9.5%44
- Pushed through a verification or sign-in5.6%26
- The plan dropped or changed what the guide said5.0%23
- Tapped while the screen was loading0.6%3
Mostly agent or prompt work, not a guide fix.
Right guide, followed, still failed
- Couldn't be done here (no data, ambiguous, blocked)98.0%146
- The task's wording left success undefined1.3%2
- The app, device or network failed0.7%1
Flows failing most
Over every reviewed day. Click a flow to list its failed runs.
| Flow | Failed | Whose fault | Top issue | Next |
|---|---|---|---|---|
| 88 | guide 0.0% · model 10.2% | Couldn't be done here (no data, ambiguous, blocked) | Agent: see the root step | |
| 60 | guide 0.0% · model 90.0% | Another wrong action or early stop | Agent: see the root step | |
| 36 | guide 0.0% · model 97.2% | Repeated actions that changed nothing | Agent: see the root step | |
| 36 | guide 5.6% · model 69.4% | Another wrong action or early stop | Agent: see the root step | |
| 32 | guide 0.0% · model 93.8% | Repeated actions that changed nothing | Agent: see the root step | |
| 31 | guide 0.0% · model 96.8% | Another wrong action or early stop | Agent: see the root step | |
| 25 | guide 0.0% · model 100.0% | Repeated actions that changed nothing | Agent: see the root step | |
| 24 | guide 0.0% · model 95.8% | The plan dropped or changed what the guide said | Agent: see the root step | |
| 17 | guide 5.9% · model 88.2% | Another wrong action or early stop | Agent: see the root step | |
| 15 | guide 0.0% · model 86.7% | Another wrong action or early stop | Agent: see the root step |