Loading ToolBench
Server has 20 tools with mixed quality. Tool names follow verb_noun convention (good), but descriptions vary widely in completeness. Most tools lack detailed parameter descriptions and output schemas. Critical issues: many parameters missing type definitions or descriptions; no documented return schemas; error handling guidance absent; some tools inferred rather than explicitly visible in source. Average description length ~120 chars (below 194 baseline). Parameter documentation is sparse, most tools show input schemas but lack per-parameter descriptions in the source code visible.
Atomic application of a full layout (zones + crossers + features)
Create a BIDIRECTIONAL area transition for adventure modules in a single call. Places a useable blue shaft of light (plc_solblue) AND a landing waypoint at BOTH positions simultaneously, guaranteeing the light and waypoint in each area are at the exact same coordinates. The A→B light's OnUsed script plays VFX_FNF_SUMMON_MONSTER_2 and jumps the PC to the waypoint in B. The B→A light does the same back to A. Both scripts are compiled and dialog resrefs are returned. Tag constraints: max 11 chars (resrefs are 'a_at_<tag>' and 'a_rt_<tag>', both max 16 chars).
Find guaranteed walkable coordinates in an area region
Procedural layout generation (BSP rooms, corridors, features)
List solver-compatible feature groups for a tileset + terrain
Create a new creature blueprint (.utc). Optionally clone from a base game or module creature and apply overrides.
14 of 20 tools have minimal or missing descriptions (under 50 chars). Examples: 'Find guaranteed walkable coordinates in an area region' (adventure_find_walkable), 'Procedural layout generation (BSP rooms, corridors, features)' (adventure_generate_layout). These lack context for when/why to use them and what they return.
No documented output/return schemas visible in source. Tools like get_area_creatures, get_area_doors, get_dependency_graph lack explicit documentation of what fields they return. LLMs cannot plan downstream calls or extract data without knowing response structure.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Create a new UTE encounter blueprint
Create a new UTM store blueprint with inventory
Find resources that are never referenced by any other resource.
List all creatures placed in a specific area with their stats.
Get full details for an area including properties, lighting, weather, and music.
List all doors in a specific area with lock status and destinations.
List all items placed on the ground in a specific area.
List all placeables in a specific area.
Analyze creature difficulty and item values per area.
Show resource dependency graph. Resources depend on other resources via script references, conversation assignments, etc.
Get a high-level summary of the entire module.
List all areas in the module with dimensions and object counts.
Set event scripts on an area's ARE file. Pass empty string to clear a script.
Check for broken references across the module (missing scripts, dialogs, items, door links).
Parameter descriptions missing or generic. adventure_create_transition has detailed param docs, but most tools (e.g., get_area_details, set_area_scripts) show parameter names and types but lack descriptions explaining what each controls or valid ranges.
No error handling guidance. Tools lack descriptions of failure modes, recovery steps, or actionable error messages. E.g., validate_module should explain what 'broken references' means and how to fix them. Per pattern:recovery-guide, errors must tell the LLM what to do next.
Destructive tools (adventure_create_transition, set_area_scripts, create_creature_blueprint) lack confirmation or dry-run support. Per pattern:confirmation-request, irreversible operations should support preview or confirmation to prevent agent mistakes.