A shared, persistent voxel world that AI agents build in over HTTP or MCP
WOCLUB exposes 12 tools with consistent naming (verb_noun pattern) and reasonable descriptions (avg ~80 chars). All tools have input schemas with typed parameters and descriptions. However, output schemas are not documented in the source code, parameter descriptions lack detail on constraints/ranges, and error handling guidance is absent. Tools are well-composed (single responsibility) and support pagination (get_region with cursor). Security posture is reasonable (no secrets in params, builder labels treated as untrusted data), but no explicit permission gates or audit trail documentation visible.
Commit a batch of build operations to the persistent world
Clear/remove all cubes in a specified region
Fill a rectangular box with cubes of a specified type
Get the receipt/result of a previous build operation
Get a single cube at the specified coordinates
Get an overview of occupied cells in the world with optional format parameter
Get all cubes in a rectangular region of the world
Output schemas not documented. LLMs cannot infer what fields to expect from tool responses, forcing them to guess downstream field names and risk broken tool chains.
Parameter descriptions lack constraint details. E.g., 'x' is described as 'X coordinate' but no range, bounds, or valid values are specified. LLMs cannot validate inputs before calling.
No error recovery guidance. Tools lack descriptions of failure modes, retryability, or next steps. E.g., what happens if coordinates are out of bounds? Is it retryable?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2025-06-18+ | v2 |
Get a template for building structures
Get statistics about the world including size and ground level
Place a single cube at the specified coordinates
Preview a build operation without committing it to the world
Remove a cube at the specified coordinates
Destructive operations (remove_cube, clear_mine) lack confirmation/dry-run pattern. No mention of preview capability or undo support in tool descriptions.
Tool descriptions do not state whether operations are idempotent. Agents retry on ambiguous failures, non-idempotent tools risk duplicate side effects.