Empty role input can make a salary tool return confident but wrong benchmarks

A developer warns that an AI agent using a salary benchmark tool can omit the requested role, receive a plausible answer for a default accounting role, and respond as if it addressed the user's job. Because the tool's input schema only requires a mode field, a request can pass validation without specifying whose salary was wanted, so a completed lookup is a weak signal of correctness. The proposed fix is to have the agent extract the role and market, require an explicit role and country, and ask the user when no role is given. Validation should run in code before any Actor call, and it must reject blank or whitespace-only input rather than relying on prompts alone. The example code and schema are synthetic and illustrate a validation boundary, not salary accuracy.