Knowledge Graph

This caches codebase analysis results in a JSON file so you don't have to re-explore the same code over and over. It stores discovered patterns, components, and APIs in docs/specs folders, then lets you query or validate against that cached knowledge when generating tasks or checking dependencies. Most useful when you're running spec-to-tasks repeatedly or need to verify that required components actually exist before implementing features. The validation is only as good as your last update though, so stale graphs can give you false confidence. Think of it as a persistent memory layer for agent discoveries that multiple team members can reference instead of each person rediscovering the wheel.