PromptLint MCP server has 2 tools with basic schemas and descriptions, but significant gaps limit production readiness. Both tools have descriptions (lint_prompt: 95 chars, auto_fix_prompt: 95 chars) and input schemas with types, but descriptions are below the 194-char baseline and lack actionable context. Parameters lack descriptions explaining constraints or expected formats. No output schema documentation. Error handling is absent, tools return unstructured strings with no recovery guidance. No tool annotations (readOnlyHint/destructiveHint). The 'magic' and 'use_ai' boolean flags lack explanation of prerequisites (LLM configuration) or failure modes. No pagination, no structured output, no per-item error reporting.

• Add parameter descriptions: 'use_ai: Enable LLM-based semantic analysis (requires OPENAI_API_KEY or LLM_API_KEY env var; slower but detects logical flaws). Defaults to false for offline analysis.' Similar detail for 'magic' parameter.
• Document output schema as structured JSON: {score: number, rating: string, issue_count: number, issues: [{severity: string, message: string, suggestion?: string}]}. Return JSON instead of markdown strings.
• Add tool annotations: lint_prompt with readOnlyHint=true; auto_fix_prompt with destructiveHint=true (modifies prompt content).
• Expand tool descriptions to 150 - 250 chars: 'Lint an AI prompt for structural issues, ambiguity, missing constraints, and logical flaws. Returns a quality score (0 - 100) and actionable suggestions. Use this before shipping prompts to production.' Similar for auto_fix_prompt.
• Implement error handling: catch exceptions, return structured errors with recovery hints (e.g., 'LLM unavailable: use_ai=false for offline analysis' or 'Invalid prompt: empty or >10000 chars').