got-controller

Graph of Thoughts orchestration coordinates generation, aggregation, refinement, and scoring operations across parallel agents. This Skill manages the GoT graph state, executes these operations in sequence, and applies KeepBestN pruning to synthesize high-quality research outputs while maintaining reproducibility logs.

Generate(k) creates multiple candidate paths; Aggregate(k) combines insights across paths; Refine(1) improves the best result; Score evaluates quality. This workflow enables dynamic pruning and parallel exploration, reducing manual overhead in complex research synthesis.

Yes. This Skill applies GoT-based orchestration to complex, multi-agent research topics across domains. It maintains centralized state coordination and operation logs, making it suitable for interdisciplinary research requiring parallel exploration and synthesized outputs.

The Skill persists a GoT graph state and comprehensive operation logs throughout execution. These logs document each Generate, Aggregate, Refine, and Score step, enabling you to reproduce research paths and audit decision points in your workflow.

KeepBestN removes lower-scoring branches after each operation cycle, reducing graph complexity while preserving the highest-quality exploration paths. This dynamic pruning accelerates synthesis without sacrificing result quality in multi-agent research orchestration.

This Skill functions as a centralized controller and integrates with other Claude Code Skills to build end-to-end research workflows. It handles orchestration and state management; complementary Skills handle domain-specific tasks upstream and downstream.