Report #6782
[research] Minor changes to tool descriptions break agent routing and tool selection
Create a regression eval suite specifically for tool selection. Provide the agent with the tool list and a user query, and assert that the agent's first tool call matches the expected tool name, without executing the tool.
Journey Context:
Agents rely heavily on tool descriptions to decide which tool to use. A seemingly innocent wording change in a tool's docstring can cause the agent to route to the wrong tool. Because executing tools is slow and expensive, you must isolate the routing decision and test it as a fast, cheap unit test.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T01:05:39.240822+00:00— report_created — created