myrmidon-swarm

To orchestrate a tiered AI swarm, this Skill decomposes complex tasks into sub-tasks and assigns them to Opus, Sonnet, and Haiku model tiers. It coordinates execution through a five-phase workflow with wave-based parallelism and dependency management.

Decomposing architectural decisions across AI model tiers is best handled by delegating high-level planning to Opus, standard execution to Sonnet, and granular tasks to Haiku. This hierarchical swarm approach ensures phased execution and traceable results for complex software engineering tasks.

Yes, you can use this phased AI workflow for large documentation efforts and code analysis. The orchestrator applies a five-phase workflow—Plan, Test, Implement, Package, Cleanup—ensuring strict safety constraints prevent unintended changes during complex task execution.

Wave-based parallelism in AI task orchestration works by executing grouped sub-tasks concurrently across assigned model tiers while managing dependencies between waves. This structured approach provides clear phase governance and ensures coordinated outcomes for multi-step planning.

No specific dependencies are required to run this tiered AI agent swarm. The Skill operates as a frontmatter-driven framework, providing built-in structured prompts, agent role definitions, and safety constraints to govern the swarm execution.

A limitation of using a hierarchical AI swarm is the strict safety constraints enforced to prevent unintended changes, which may restrict autonomous execution in highly dynamic environments. The phased workflow also requires tasks to fit within the defined Plan, Test, Implement, Package, and Cleanup structure.