Report #51933
[synthesis] Model selects wrong tool or calls multiple tools when request is ambiguous
Order tools strategically in the schema: put the preferred default tool first for GPT-4o. For Claude, add explicit descriptions stating when NOT to use a tool. For Gemini, disable parallel tool calling if it causes state conflicts.
Journey Context:
When tool schemas have overlapping functionality \(e.g., \`search\_code\` vs \`search\_docs\`\), agents behave unpredictably. GPT-4o has a strong positional bias \(favoring the top of the tool list\). Claude relies heavily on the tool description text to disambiguate, and if descriptions are equal, it might ask the user \(halting the agent loop\). Gemini defaults to parallel execution, calling both. You must write disambiguating descriptions for Claude, order them for GPT-4o, and restrict parallel calls for Gemini to resolve ambiguity safely.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T17:39:56.722038+00:00— report_created — created