Loading ToolBench
MCP server for rolling dice using standard RPG dice notation
Single tool with clear naming and functional schema. Descriptions are adequate but could be more LLM-optimized. Parameter descriptions are present and specific. Output schema is not explicitly documented in code. Error handling is basic but present. The tool follows core pattern disciplines but lacks some production-grade refinements around output structure documentation and error guidance.
Roll dice using standard dice notation (e.g., "3d6", "d20", "2d10"). Supports rolling multiple dice of various sizes.
Output schema is not explicitly documented. The tool returns text content with both formatted and raw JSON representations, but LLMs cannot introspect the structure of DiceRollResult without seeing the schema in the tool definition.
Error messages provide raw error text but lack actionable recovery guidance. When notation parsing fails, the error states the problem but does not guide the LLM to retry with corrected syntax.
Tool description mentions examples ('3d6', 'd20', '2d10') in free text rather than as enum constraints or pattern validation. This risks LLMs treating these as the only valid options or inferring formats incorrectly.
No input validation constraints (min/max dice count or sides) are exposed in the schema descriptions. Validation rules exist in code (max 1000 dice, max 1,000,000 sides) but are not communicated to the LLM via parameter metadata.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 56 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 59 | - | v1 |