Loading ToolBench
AI Unity development harness that integrates workflows, Skills, CLI/Runtime tools, and validation loops for AI coding assistants.
AIBridge exposes 7 tools via HTTP, but definition quality is severely compromised. Tool names lack action verbs (health, commands, results, cleanup, artifacts, discovery, code_index_query). Descriptions are present but generic and lack LLM-optimized guidance on WHEN to use each tool or WHAT happens (especially for write operations). Input schemas exist for most tools but are minimal, parameters lack type constraints, enums, ranges, or validation rules. Output schemas are completely undocumented. Error handling is absent from tool definitions. The 'commands' tool accepts arbitrary command strings with no validation, format guidance, or safety constraints. No tool declares permissions, idempotency, or destructiveness. This server reads like a raw HTTP API wrapper, not an LLM-optimized tool suite.
GET /aibridge/artifacts/{filename} - Retrieve screenshot artifacts generated by the runtime
DELETE /aibridge/commands/{commandId} - Clean up and forget a command from the runtime
Query the Code Index daemon for C# declaration-name lookup and declaration location information
POST /aibridge/commands - Submit and execute commands on the AIBridge runtime
UDP LAN discovery protocol for AIBridge runtime instances - responds to discovery requests with runtime metadata
GET /aibridge/health - Returns health status of the AIBridge runtime
GET /aibridge/results/{commandId} - Retrieve results of a previously executed command
Tool names lack action verbs. 'health', 'commands', 'results', 'cleanup', 'artifacts' do not start with verbs (get_, list_, create_, delete_, search_). LLMs cannot infer intent from these names alone.
No output schemas documented. Tools return HTTP responses but LLMs have no visibility into response structure, field names, or types. Agents cannot plan downstream calls or extract required data.
'commands' tool accepts arbitrary command strings with zero validation, format guidance, or safety constraints. No enum, pattern, or description of valid command syntax. Invites injection attacks and hallucinated commands.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
No error handling guidance in tool definitions. No indication of retryable vs fatal errors, no recovery suggestions, no actionable error messages. Agents cannot self-correct on failure.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). 'cleanup' and 'commands' are destructive but unmarked. Agents cannot reason about side effects or retry safety.