Extensible agent platform for Ollama with MCP server support, tool registry, chat sessions, and skill management
Only 1 tool visible in source (echo). Tool has a basic description and input schema with a single string parameter. However, the description is trivial (10 chars: 'Echoes the given text back, verbatim.'), parameter lacks a description, and the tool is READ_ONLY with no real agent value. The server appears to be a demo/test harness (modelbench) rather than a production agent tool suite. MCP client integration code is present (pkg/mcpclient/tool.go) showing adapter pattern for remote tools, but the server itself exposes only a toy echo tool. No evidence of proper tool composition, error handling guidance, or LLM-optimized descriptions.
Echoes the given text back, verbatim.
Tool description is trivial (10 chars). Rubric requires 10 - 1024 chars with context on WHAT, WHEN, and WHY. 'Echoes the given text back, verbatim.' provides no agent selection guidance.
Parameter 'text' has no description. Rubric requires every parameter to have a non-empty description explaining what it controls. LLMs cannot infer meaning from names alone.
Only 1 tool exposed by the server (echo). This is a test/demo harness, not a production agent tool suite. The server's value proposition is unclear, echo has no real agent utility.
No error handling guidance. Tool has no documented error cases, recovery paths, or actionable error messages. Rubric requires error responses to tell the LLM what to do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2025-06-18+ | v2 |
No output schema documented. Rubric requires documentation of what the tool returns so LLMs can plan downstream calls. Echo presumably returns a string, but this is not explicit.