Four tools with clear verb-noun naming (list_*, approve_*, reject_*) and solid descriptions (120-180 chars each). All have proper input schemas with enums and required fields. However, output schemas are completely undocumented, LLMs cannot plan downstream calls or extract structured data. Parameter descriptions are minimal (10-30 chars), lacking format guidance and constraints. No error handling guidance. Missing idempotency hints for write operations. Descriptions are adequate but not LLM-optimized per pattern:tool-description baseline (194 chars avg, 50-200 target).

• Document output schemas for all tools. E.g., list_pending_candidates should return {candidates: [{id: int, category: string, statement: string, confidence: float}], total: int}.
• Expand parameter descriptions to 50-100 chars with format/range guidance. E.g., 'candidate_id (integer, 1-999999): Unique identifier from list_pending_candidates output.'
• Add pagination support: limit (1-100, default 20) and offset/cursor parameters to list_* tools. Return total_count and next_cursor in responses.
• Add error handling guidance in descriptions: 'Returns 404 if candidate_id not found. Returns 409 if already approved/rejected. Retryable on 5xx errors.'
• For write operations, add idempotency hints and optional dry_run parameter. E.g., approve_candidate with dry_run=true returns what would happen without committing.

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

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