Two tools with complete schemas and reasonable descriptions. calculate_work_hours has detailed parameter documentation and enum constraints (federal|california|alaska|colorado|nevada, rounding options). time_and_a_half is simpler but well-defined. Both tools lack output schema documentation, LLMs cannot predict response structure. Descriptions are adequate (194 - 250 chars, within baseline 34 - 392 range) but lack actionable context on when to use each tool vs. the other. No error handling guidance, no recovery paths for invalid inputs. Parameter descriptions are present but some lack format/range details (e.g., 'ISO date yyyy-mm-dd' is good; 'HH:mm, 24-hour' is good). Missing: output field documentation, error classification, and composition guidance.

• Document output schemas for both tools. Example for time_and_a_half: {regularRate: number, overtimeRate: number, overtimeHours: number, overtimePay: number, source: string}. This enables LLMs to extract and chain results.
• Add error handling guidance: 'Invalid date format: expected yyyy-mm-dd. Invalid hourly rate: must be ≥ 0. Invalid ruleset: must be one of federal|california|alaska|colorado|nevada.' Include recovery hints.
• Expand tool descriptions to clarify composition: 'Use calculate_work_hours for multi-day timesheets with jurisdiction-specific rules. Use time_and_a_half for simple 1.5x rate calculations on a single overtime amount.'
• Document the default behavior of unpaidBreakMins=0 and when to override it (e.g., 'Set to 30 if employee took a 30-minute unpaid lunch break during the shift').
• Add idempotency note: 'Both tools are read-only and idempotent, repeated calls with identical inputs always return identical results. Safe to retry on transient failures.'