dream-planning

To decompose complex software engineering tasks for AI agents, you break work into structured, context-isolated plans using magnitude routing and plan metadata, enabling predictable decomposition and disciplined workflow for multi-step planning.

Magnitude routing in AI agent workflows is a formal protocol that directs work into parallelizable, isolated units. It uses plan hierarchy and task definitions to enforce context isolation and ensure safer collaboration across modules.

You enforce context isolation when dispatching subagents by applying sibling firewall constraints through _overview.md frontmatter and directory naming, ensuring each AI agent operates within clear boundaries and does not cross-module route unintentionally.

The difference between a System Plan (SP) and a Procedure Plan (PP) lies in their structured plan types with naming conventions. SP and PP definitions support scalable architectures by categorizing engineering atomic modules within the plan hierarchy.

Yes, you can use this decomposition protocol without dependencies. The DREAM planning protocol requires no external dependencies to break work into parallelizable units and enforce sibling firewall constraints for AI-driven teams.