MCPKernel has 10 tools with complete schemas and descriptions, but quality is uneven. Core security tools (mcpkernel_scan_tool, mcpkernel_validate_policy, mcpkernel_check_taint) have domain-specific descriptions (60-90 chars) but lack actionable guidance on when to use them vs alternatives. Example tools (greet, add, echo, system_info) are trivial and dilute the evaluation. Parameter descriptions are present but minimal (10-30 chars). No tool declares error recovery paths or permission requirements. Output schemas are not documented. Tool names follow verb_noun convention but lack clarity on mutual exclusivity (e.g., mcpkernel_scan_tool vs mcpkernel_scan_skill both scan but different formats).

• Document output schemas for all tools. Example: mcpkernel_scan_tool should return {issues: [{type: string, severity: string, description: string}], passed: boolean}. This enables LLMs to extract and act on results.
• Expand parameter descriptions to 50-100 chars with format constraints. E.g., 'Path to YAML policy file (absolute or relative to .mcpkernel/; must exist and be readable; max 10MB).'
• Add error recovery guidance to tool descriptions. E.g., 'If policy validation fails, returns error with line number and suggestion. Call mcpkernel_doctor to check file permissions.'
• Clarify tool selection in descriptions. Prefix mcpkernel_scan_tool with 'Scan MCP tool definitions...' and mcpkernel_scan_skill with 'Scan OpenClaw SKILL.md files...' to disambiguate.
• Remove or relocate example tools (greet, add, echo, system_info) to a separate examples/ server. They add noise and confuse agents about the server's actual purpose.

Inferred effective spec: <=2025-11-25.

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