Roll dice, flip coins, and handle D&D mechanics including ability scores, advantage/disadvantage, initiative, hit dice, saving throws, and random table selection.
The Dice Roller server provides 9 tools with consistent naming (all verb_noun) and reasonable descriptions. However, critical gaps limit its quality: (1) Parameter descriptions are present but generic, many lack format constraints, ranges, or validation guidance. (2) Output schemas are entirely undocumented; responses are plain text strings, not structured objects. (3) Error handling is minimal, no categorization, no recovery guidance, and no actionable error messages. (4) Parameters like 'disadvantage' and 'advantage' use string 'true'/'false' instead of booleans, requiring LLMs to guess the expected format. (5) Tool descriptions, while adequate (average ~130 chars), lack dependency hints and preconditions. The tools themselves are well-implemented (good validation, reasonable defaults), but the interface documentation falls short of production standards.
Flip one or more coins. Returns heads/tails results.
Roll a full set of six D&D 5e ability scores using the 4d6-drop-lowest method.
Roll dice using standard notation (e.g. 2d6, 1d20+5, 4d6kh3). Supports NdS, modifiers (+/-), keep-highest (kh), and keep-lowest (kl).
Roll hit dice for healing during a short rest. Specify dice (e.g. 2d10) and CON modifier.
Roll initiative for one or more creatures. Provide comma-separated names and a shared modifier, or just a count.
Roll a d100 (percentile dice).
Pick a random item from a comma-separated list. Great for random encounters, loot, or any table.
No output schemas documented. All tools return plain-text strings; LLMs cannot parse structured results or chain tools requiring specific fields.
Boolean parameters use string 'true'/'false' instead of proper boolean type. Examples: 'disadvantage' in roll_with_advantage, 'advantage'/'disadvantage' in saving_throw. Forces LLMs to guess string format and introduces parsing errors.
Parameter descriptions lack format constraints and validation guidance. 'notation' accepts free-form strings with no regex pattern or example format in description. 'modifier' and 'count' lack min/max bounds in descriptions. 'items' in roll_table lacks guidance on delimiter or max item count.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 58 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 42 | - | v1 |
Roll a d20 with advantage or disadvantage (D&D). Set disadvantage='true' for disadvantage, otherwise rolls with advantage.
Make a D&D saving throw against a DC. Specify the DC, your modifier, and advantage/disadvantage.
Error handling returns plain error messages (e.g. 'Invalid notation: ...') with no categorization, recovery guidance, or actionable next steps. LLMs cannot distinguish retryable errors from user-fixable ones.
roll_percentile has minimal description ('Roll a d100 (percentile dice).'), under 50 chars, lacking context for when to use it vs roll_dice.
Parameter 'names' in roll_initiative defaults to empty string but description doesn't clarify expected delimiter (comma-separated) or max count. LLMs may pass invalid formats.