Server defines 5 tools with complete input schemas and descriptions. Naming follows verb_noun convention (get_*, check_*). Descriptions are adequate (80-180 chars) but lack LLM-optimized guidance on WHEN to use each tool vs. alternatives. Parameters are typed and mostly described, but lack constraints (enums, ranges, format hints). Output schemas are not documented, LLMs cannot predict response structure. Error handling exists but is generic. No tool annotations (readOnlyHint, etc.). check_code tool has good heuristic logic but lacks documented output schema and confidence scoring guidance.

• Document output schema for each tool. Example for check_code: 'Returns {findings: [{id, title, category, wcag, level, matchIndex, line, evidence, confidence, kind}], summary: string, stats: {total, byLevel, byCategory}}'.
• Add readOnlyHint: true to all tool schemas to signal safety to agents. Update ListToolsRequestSchema response to include annotations.
• Enhance get_rule description: 'Retrieve a single rule by ID (e.g. img-has-alt) or WCAG criterion (e.g. 1.1.1). Use this to get detailed guidance on a specific violation.'
• Document check_code confidence levels in description: 'Confidence: high (pattern match), medium (heuristic), low (ambiguous). Use high-confidence findings first.'
• Add error recovery guidance: 'If query not found, call get_all_rules() to see available rule IDs, or get_category_rules() to browse by category.'

Inferred effective spec: 2025-06-18+.

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