An agent produced a wrong change. Logs show it called the right tools, but it misread the requirement. How do you classify this?
Why B is the answer
A reasoning error
Classifying root causes properly decides the fix. Here, instructions need revising, and tool access does not.
In detail
- The tools were used correctly.
- It misread the requirement, so the failure is in reasoning.
- Classify the failure before you tune anything.
Use case
An illustrative scenario, not a real incident.
The right tools ran with the right arguments, but the agent understood "archive" as "delete". That is a reasoning error, so the fix is clearer requirements.
See it in different ways
flowchart TD
A["Agent failure"] --> B{"Where did it go wrong?"}
B -->|"Misread the task"| C["Reasoning: improve instructions"]
B -->|"Wrong tool or input"| D["Tool misuse: fix tool definitions"]
B -->|"Missing dependency"| E["Environment: fix setup"]stateDiagram-v2 [*] --> Failed Failed --> Reasoning: right tools, wrong understanding Failed --> ToolUse: right tool, wrong arguments Failed --> Environment: missing dependency or access Reasoning --> Tune: instructions and context ToolUse --> Tune: tool definition Environment --> Tune: setup
flowchart TD
Q{"Where did it go wrong?"}
Q -->|"Understanding"| A0["Reasoning error"]
Q -->|"Tool arguments"| A1["Tool misuse"]
Q -->|"Setup"| A2["Environment failure"]Key point: Right tools and a wrong understanding means reasoning.
Why the other options are wrong
- A. A network failure
- Nothing about the network failed.
- C. A tool misuse failure
- The logs show the right tools were called.
- D. An environment failure
- The environment worked.