A trip expense splitter and debt settlement calculator for group travel in Taiwan, with attraction search and seeded demo data
8 tools with explicit schemas and descriptions. Naming is clear and verb-based (toCents, fromCents, fmt, splitExpense, balances, settle, searchAttractions, seedTrip). Descriptions range 50 - 180 chars, meeting the 10 - 1024 baseline. All tools are READ_ONLY. However, parameter descriptions are minimal (10 - 30 chars), lacking context on valid ranges, formats, and dependencies. Output schemas are not documented, LLMs cannot predict return structure. Error handling is absent; no guidance on invalid inputs or recovery paths. The splitExpense tool accepts a complex nested object but does not document the required structure or field constraints.
Calculate net balance per member across all expenses. Positive balance means others owe them (they paid more than their share).
Format an amount in cents as a localized currency string for Taiwan (Traditional Chinese)
Convert integer cents back to a decimal number
Search attractions from the curated Taiwan tourism database by name, city, town, or class. Returns up to limit results (default 8).
Generate a seeded demo trip with 4 members and 4 sample expenses for a 3-day Taiwan trip to Yilan, Qingjiang, and night markets. Used to populate the UI with initial data.
Greedy debt simplification algorithm that computes minimum number of transfers to settle all debts between group members
Split one expense into per-member owed amounts in cents. Supports equal, exact, and percent split modes. Remainders are distributed to earliest participants.
Output schemas not documented. LLMs cannot predict return structure for balances(), settle(), searchAttractions(), or seedTrip(). This forces agents to guess field names and types, increasing errors.
splitExpense accepts a complex nested 'exp' object but does not document required fields, valid enum values for 'mode', or constraints on shares object. LLMs cannot construct valid calls without trial-and-error.
Parameter descriptions are minimal (10 - 30 chars). 'The amount to convert to cents' and 'Currency symbol (defaults to NT$)' lack context on valid ranges, formats, or edge cases. Baseline is 72 chars average.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Convert a number to integer cents to avoid floating-point drift
No error handling or recovery guidance. Tools do not document what happens on invalid input (e.g. negative amounts, invalid currency, empty search query, out-of-range limit). LLMs have no guidance on retry or fallback.
searchAttractions returns up to 8 results by default but does not document pagination support, total count, or next_cursor. Large result sets could exhaust context; no guidance on how to fetch more results.