Server has 11 tools with basic schemas and descriptions, but significant quality gaps. Most tool descriptions are under 100 chars and lack actionable context. Parameter descriptions are minimal or missing. No output schemas documented. Error handling guidance absent. Tool names are reasonable (verb_noun pattern) but lack clarity on when to use each vs. similar tools. Sampling feature is deprecated (removed in 2026-07-28 spec). No tool annotations (readOnlyHint/destructiveHint) despite having WRITE operations. Schemas present but incomplete, many parameters lack type constraints and validation guidance.

• Add output schema documentation for all 11 tools. Specify return type (object/array), required fields, and field types. Example: lint_prompt returns {issues: [{type: string, severity: 'error'|'warning', line: number}], summary: string}.
• Add tool annotations: mark generate_agent_onboarding, ingest_pattern, generate_template with destructiveHint=true; mark lint_prompt, evaluate_prompt with readOnlyHint=true. Use MCP SDK's tool annotation API.
• Expand parameter descriptions to 50-150 chars with format/constraint guidance. Example: 'gaps: Array of gap objects from lint_prompt output, each with {type: string, description: string, severity: string}' instead of current generic description.
• Remove or refactor proactive_suggest (Sampling is deprecated). Either integrate with LLM provider's native API or replace with a deterministic tool that returns structured suggestions without LLM-to-LLM calls.
• Add error handling guidance to tool descriptions. Example: 'If lint_prompt fails due to unsupported language, try with semantic=false for rule-based linting only. Returns {error: string, retryable: boolean, suggestion: string}'.

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

• Relies on Sampling (deprecated) - integrate directly with the LLM provider API