hive-mind-advanced

Multi-agent coordination uses a queen-led hierarchical architecture where diverse agents with specialized roles (strategic, tactical, adaptive) collaborate through consensus mechanisms. This Skill implements Byzantine fault tolerance requiring a 2/3 majority to ensure robust decision-making even when some agents fail or disagree.

Collective memory is a shared, persistent knowledge base all agents access and contribute to during task execution. It uses LRU caching and SQLite WAL persistence to store learnings, patterns, and decisions, enabling agents to learn from past work and avoid repeating mistakes across sessions.

Yes. This Skill orchestrates queen-led coordination across worker agents with distinct specializations—coders, testers, architects, reviewers—executing strategic, tactical, and adaptive roles. All decisions pass through Byzantine consensus, and collective memory tracks learnings across specializations.

Byzantine consensus requires a 2/3 majority agreement before any decision is committed, tolerating up to one-third of agents being faulty, delayed, or malicious. This ensures decisions remain valid even when individual agents fail or return incorrect results.

The Skill depends on claude-flow for agent orchestration and requires initializing the queen-led hierarchy, defining worker agent roles and specializations, configuring collective memory storage (SQLite backend), and setting up session and task distribution parameters.

Yes. This Skill includes memory access pattern tracking and optimization, monitoring how agents retrieve and update collective memory. This enables performance tuning and identifying which knowledge is most valuable for future task optimization.