Single tool with no input parameters and minimal schema definition. Tool description is present (194 chars) but lacks actionable guidance on when/why to use it. No input schema validation, no output schema documentation, no error handling guidance. Tool name 'get_sysinfo' follows verb_noun convention but is generic. No parameter descriptions exist (tool accepts no params). Output is unstructured text rather than typed fields. Fails multiple critical checks: no documented output schema, no pagination support, no error recovery guidance.

• Convert text output to structured JSON: return {system: {name, node, release, version, machine, processor}, cpu: {processor, physical_cores, logical_cores}, memory: {total_bytes, available_bytes, used_bytes, utilization_percent}, disk: {total_bytes, used_bytes, free_bytes, utilization_percent}} instead of concatenated strings.
• Add output schema documentation in tool docstring: 'Returns a SystemInfo object with nested system, cpu, memory, and disk fields. All byte values are integers; percentages are 0-100 floats.'
• Enhance description with actionable context: 'Call this to retrieve current system resource metrics (CPU, memory, disk, OS info). Use before capacity planning or troubleshooting. Returns structured data suitable for monitoring dashboards. No parameters required.'
• Add error handling: wrap psutil/platform calls in try-except, return {error: 'Failed to retrieve <component> info: <reason>'} with specific component names so LLM knows what failed.
• Add tool annotations via FastMCP: use @mcp.tool(annotations={'readOnlyHint': True, 'idempotentHint': True}) to signal to agents that this is safe to call repeatedly without side effects.