Loading ToolBench
This server presents a minimal demo with two simple arithmetic/utility tools. Both tools have basic descriptions and partial schemas, but lack the rigor expected in production. Naming is acceptable but descriptions are extremely brief (under 20 chars for roll_dice after stripping boilerplate). Parameter descriptions are missing entirely for the 'sides' parameter in roll_dice. Output schemas are not documented. No error handling guidance, no parameter constraints or validation hints. The tools are trivial (random integer, addition) and show no sophisticated composition patterns. This is a learning/demo server, appropriate for that scope, but would not meet production quality standards.
Add two numbers.
Roll a dice with a given number of sides.
roll_dice: Parameter 'sides' lacks description. LLMs cannot infer that sides must be >= 1, what the default of 6 means, or the range of valid values.
Both tools lack documented output schemas. LLMs do not know the structure of return values (is roll_dice returning an int or {roll: int}?). This breaks downstream tool chaining and forces LLMs to guess.
roll_dice description ('Roll a dice with a given number of sides.') is 47 characters, minimal and lacks context. No mention of return range (1 to N), when to use it, or edge cases. Production baseline is 50-200 chars with explicit context.
add: Tool name is too generic. 'add' could mean string concatenation, list merging, or numeric addition. Better name: 'add_numbers' or 'sum_integers' to clarify intent and reduce ambiguity when tool count grows.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 41 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
No error handling or validation guidance. What if sides < 1? What if side effects fail? LLMs receive no actionable recovery instructions (e.g., 'sides must be >= 1; try again with a valid value').
No input validation hints. The 'sides' parameter has a default but no min/max bounds documented in the description. 'sides=1000000' could be passed; LLMs have no guard rails.