What This Looks Like
The agent needs information from a tool result to decide which tool to call, but it must choose a tool before it can get that information. The workflow may stall, guess, call the wrong tool, ask unnecessary questions, or loop between possible tools without a valid selection rule.
Why It Matters
Tool selection needs enough input to be decidable. When the required evidence depends on the tool choice itself, the agent can become trapped in a circular dependency. This creates unreliable tool use and makes failures look like model indecision rather than a broken selection structure.
Structural Signal
The tool-choice decision depends on information that is only available after the tool-choice decision. The issue is not one bad tool call; it is a circular dependency in the tool selection process.
Common Triggers
- Tool descriptions require knowing the result before selecting the tool
- The workflow lacks a discovery or metadata step
- Multiple tools appear valid until one is called
- Required routing fields can only be generated by downstream tools
- The agent is not allowed to ask for missing selection inputs
- Tool selection rules are described in terms of unavailable state
When to Use This Issue
Use this Issue when an agent cannot choose the correct tool because the information needed to choose is only available after calling a tool.
When Not to Use This Issue
Do not use this Issue when the required tool is available and the agent simply picks the wrong one. Do not use it when the tool exists but required inputs are missing outside a circular selection dependency.