Single tool with minimal definition quality. Tool name 'aggregate_votes' lacks action verb prefix (should be 'aggregate_preference_votes' or similar). Description is present (89 chars) but generic and lacks WHEN/WHY guidance. Input schema visible but parameters lack individual descriptions, only high-level array descriptions provided. No output schema documented. No error handling guidance. No parameter constraints (enums, ranges, formats). Tool does not follow verb_noun convention clearly and lacks LLM-optimized structure.

• Add individual parameter descriptions: 'candidates' (array of strings, 2-100 items, unique identifiers for options to rank), 'ballots' (array of objects with voter_id: string, ranking: ordered array of candidate IDs, weight: float 0.1-10.0 credibility multiplier). Document ballot structure explicitly.
• Document output schema: 'Returns object with ranking (array of [candidate_id, borda_score] tuples sorted descending), borda_scores (dict mapping candidate_id to weighted score), condorcet_winner (string or null if no candidate beats all others pairwise).'
• Expand tool description to 150-200 chars: 'Aggregates weighted preference ballots using Borda count voting. Computes positional scores weighted by voter credibility, detects Condorcet winners via pairwise comparison, and returns ranked candidates. Use when combining multi-agent or stakeholder preferences with varying trust levels.'
• Add error handling: return structured errors for invalid inputs (e.g., 'Invalid ballot: ranking contains unknown candidate X. Valid candidates: [...]', 'Insufficient ballots: need at least 1 ballot per candidate for reliable ranking'). Include recovery hints.
• Rename tool to 'compute_ranked_preferences' or 'aggregate_preference_votes' to lead with action verb and improve discoverability.