Nexarion exposes 7 tools that bridge MCP and A2A protocols. Critical gaps: (1) Most tools lack input schemas entirely, tools/list, resources/list, prompts/list have no visible schema definitions. (2) Descriptions are minimal (10-50 chars) and lack actionable context for LLM selection. (3) No output schemas documented. (4) tools/call accepts generic 'arguments' object with no structure. (5) No error handling guidance, recovery hints, or validation rules. (6) Tool names use slash notation (tools/list) rather than verb_noun convention, making intent ambiguous. (7) No parameter descriptions beyond bare names. This server prioritizes protocol bridging over agent usability.

• Rename tools to verb_noun convention: list_tools, call_tool, list_resources, read_resource, list_prompts, get_prompt. This matches LLM expectations and production baselines.
• Add 50-200 character descriptions to each tool explaining WHAT it does, WHEN to use it, and what it returns. E.g., 'List all MCP tools exposed from connected A2A agents. Call this first to discover available operations before invoking tools/call.'
• Document input schemas for all tools. For tools/call, replace generic 'arguments' with a structured schema: {type: 'object', properties: {tool_name: {type: 'string'}, params: {type: 'object', description: 'Tool-specific parameters'}}}.
• Add output schemas to tool definitions. E.g., tools/list should return {type: 'array', items: {type: 'object', properties: {name: {type: 'string'}, description: {type: 'string'}, input_schema: {type: 'object'}}}}.
• Add error handling guidance: 'If tool not found, call list_tools() first to discover available operations. If call fails, check tool_name and arguments match the schema returned by list_tools.'
• Add parameter descriptions. E.g., resources/read uri param: 'The URI of the resource to read (format: scheme://path, e.g., file://config.json or http://api.example.com/data).'
• Document pagination for list_* tools. Add limit and offset parameters with defaults (limit=20, offset=0) and return total_count so agents can iterate.

Inferred effective spec: 2026-07-28+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry