Loading ToolBench
This is a minimal starter template with only 1 tool. The 'add' tool has a clear verb-noun name and basic description, but lacks depth expected for production use. Tool description is concise but minimal (18 chars). Parameter descriptions are present in the schema but very brief (9-12 chars each). No output schema is documented, no error handling guidance, no composition patterns beyond the single tool. The resource endpoint ('greeting://{name}') is also present and functional but similarly minimal. This is typical of a tutorial/starter repo, adequate for learning but not suitable for production agent deployment.
Add two numbers
Tool description is only 18 characters ('Add two numbers'). Baseline for production tools is 34 - 392 chars (p10=34, p90=392). Description lacks context on WHEN to use this tool, WHAT it returns, and any prerequisites.
Parameter descriptions in schema are extremely brief ('First number', 'Second number', 12 - 15 chars each). Baseline is 72 chars. These lack information about constraints, valid ranges, or format expectations. LLMs cannot infer intent from bare type+name alone.
No output schema is documented for the 'add' tool. Callers must infer that it returns an integer. LLMs need explicit documentation of return types and structure to plan multi-step workflows.
No error handling or recovery guidance. If called with invalid input (non-integer, overflow), there is no documented error response or hint for the LLM on how to recover.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 50 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 28 | - | v1 |
Parameters lack constraints (min, max, range). Numeric parameters should specify bounds (e.g., 'integers in range -1e6 to 1e6') to prevent LLMs from passing absurd values.