hive-mind-advanced

Multi-agent coordination uses a queen-led architecture where one agent directs specialized workers through hierarchical roles. This Skill orchestrates agent collaboration, consensus decision-making, and shared memory so all agents stay aligned on project goals and share knowledge continuously.

Byzantine consensus is a fault-tolerant voting mechanism requiring a 2/3 majority approval, designed to handle faulty or unreliable agents. It ensures robust decisions even when some agents fail or behave unexpectedly, critical for mission-critical multi-agent systems.

Collective memory uses SQLite WAL persistence with LRU eviction to store and retrieve shared knowledge across all agents. This enables continuous learning—agents access prior decisions and context, avoiding redundant work and improving decision quality over time.

Yes. A queen agent directs specialized workers (coders, testers, architects) through hierarchical roles while consensus validates major decisions and collective memory tracks all project context. This supports large-scale, ongoing projects requiring coordination across research, development, QA, and analytics.

Three consensus algorithms are supported: Majority (simple vote), Weighted (votes with influence scores), and Byzantine (fault-tolerant 2/3 majority). Choose based on your fault tolerance needs and whether agents have different expertise levels.

No specialized infrastructure is required. Session-based management and SQLite persistence work in standard environments. The Skill handles queen-worker role assignment, memory eviction, and consensus logic internally—start with the init command and spawn agents as needed.