Loading ToolBench
A Model Context Protocol server that exposes the AdventureEngine as 18 tools for AI agents to autonomously playtest and debug a retro-terminal LLM text adventure game with engine-authoritative game state
17 tools with consistent naming (verb_noun pattern) and basic descriptions, but significant gaps in schema completeness and parameter documentation. All tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization), and most parameters lack type definitions in visible schemas. No output schemas documented. Error handling is absent. Naming is clear and action-oriented, but descriptions are too terse to guide LLM tool selection effectively.
Add an entry to the adventure's memory system
Delete an adventure and its associated data
Execute a barter/trade transaction
Find a route between two rooms in the spatial map
Get the current state of the barter/trading system
Get the current game state
Get diagnostic information about the engine and session
Get the spatial map of rooms and connections
List all available adventures
Descriptions are too terse (10-50 chars vs. 50-200 char baseline). LLMs cannot determine when to select tools or distinguish similar operations (e.g., set_location vs. set_score both say 'Set the...'). Descriptions lack WHEN to use, prerequisites, or side effects.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. Agents cannot chain tools (e.g., new_adventure returns what? An adventure_id? A full state object?).
Destructive operations (delete_adventure) lack confirmation or dry-run support. No error guidance for recovery. Agents cannot safely retry or understand consequences.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Load an existing adventure by ID
Create a new adventure session
Query the adventure's memory system for relevant context
Send a player command to the adventure engine
Set the player's current location
Set the move counter
Set the player's score
Undo the last turn in the adventure
Parameter descriptions are missing or generic. 'The adventure ID to load' does not explain format (UUID? string? numeric?). 'The player command to execute' does not list valid commands or constraints. LLMs will hallucinate invalid values.
No error handling or recovery guidance. Tools do not indicate what to do if a command fails, an adventure is not found, or a barter is invalid. Agents cannot self-correct.