A maker builds an agent whose task instructions explicitly say to look up information with a specific tool — for example, finding attractions for a trip-planning assistant — but the agent answers without ever calling it.
Symptom
A task's step-by-step Instructions explicitly mention using a particular tool, and that tool is added to the task's Tools list, but at runtime the agent skips it — it answers directly (often incompletely or inaccurately) instead of calling the tool. Checking the session trace confirms the tool never appears in the execution flow at all.
Cause
The model doesn't decide which tool to call from your task instructions alone — it also reads each tool's own description, configured on the tool itself (in its Context tab, alongside the tool's name). That description is what the model actually reasons over to decide when a tool is relevant.
If a tool's description is missing, vague, or doesn't clearly say what the tool does and when to use it, the model can leave it out of its plan even when your task instructions reference it by name — the instructions point at the tool, but the tool itself doesn't explain itself well enough for the model to pick it.
Fix
Find the tool where it was added. For a tool added to a task, open Tasks, edit the task and select the tool in its Tools list. For an agent-level tool, open Tools and choose Configure.
Open the Context tab and review its name and description.
Rewrite the description so it clearly states what the tool does, when to use it, and (in plain terms) what it returns — a specific description like "Finds tourist attractions and activities based on user preferences and location; use when the user asks for things to do or places to visit" works far better than a one-word label or a blank field.
Go back to the task under Tasks and confirm its Instructions explicitly reference the tool by the situations where it should be used, not just a generic "use tools when needed."
Save your changes, then use Test to run a query that should trigger the tool.
To verify, open the trace timeline for your test session and confirm the tool now appears in the execution flow with the expected input and output, and that the agent's final response incorporates that tool's output.
Prevention
A tool without a clear description is effectively invisible to the model's own tool-selection reasoning, no matter how explicitly your task instructions mention it by name. Treat every tool's description as mandatory, working documentation the model relies on at runtime — not an optional label — and use the trace timeline to confirm a tool is actually being selected rather than assuming task instructions alone are enough.
Note: Task instructions and tool descriptions serve different purposes — instructions tell the agent what to accomplish, but the tool description is what the model reasons over to decide whether a given tool applies to the situation in front of it.