graph-skills

Reduce context usage and cost in Claude Code workflows by using graph-based orchestration to automatically route tasks across Haiku, Sonnet, and Opus, leveraging fast models for simple steps and powerful models for complex steps.

Graph-based workflow orchestration is a method that uses a topological graph engine to route tasks across different Claude models, optimizing context efficiency by assigning explicit node inputs, outputs, and dependencies to the appropriate model.

Build a repository analysis pipeline by running the example-repo-summary.ts script, which executes a graph that routes repository analysis tasks across Claude models to optimize context usage and cost.

You can use Haiku, Sonnet, and Opus in the same Claude Code task through a model-router with cost-aware heuristics that transparently routes nodes to the optimal model per task within a graph-based framework.

Use graph-based skills instead of standard subagents when building complex, multi-step data processing workflows that require explicit node inputs, outputs, dependency handling, and cost optimization across multiple Claude models.