-
Designing agent tools as a deliberate contract
Treat each tool exposed to an agent as its own design problem: namespace related tools under a shared prefix, consolidate multi-step workflows into one tool where that matches how the task is actually done, keep responses terse and high-signal (pagination/truncation, semantic names over raw IDs), and iterate against real evals.
Why this verdict?
Anthropic's tool-design guide frames tools as a contract between deterministic software and a non-deterministic reasoner, and reports that small wording and consolidation changes produced outsized gains on real evals -- consistent with their broader finding that ambiguity in a tool list becomes ambiguity in agent behavior.