rise-pde-session-multi-agents-v2

What problem does it solve?

It orchestrates an end-to-end RISE (Reverse-engineer → Intent-extract → Specify → Export) workflow for implementing, documenting, and integrating changes using PDE decomposition plus parallel multi-agent execution.

Core Features & Use Cases

• PDE-driven decomposition and STC charting: Decomposes the user’s intent into a structured action stack and generates an STC (structural tension chart) from the PDE decomposition for human review.
• Enforced review window with wake-up execution: Runs an enforced bash sleep (sleep 1500) to give humans time to review and edit the authoritative .pde/ files before execution resumes.
• Parallel deep-search, validation, and execution by direction: Uses SOUTH deep-search agents, a WEST cross-reference validator, and NORTH executor agents with file-ownership boundaries and kinship/rispecs awareness.
• Supports multi-part cascading missions: Detects and executes “Part 1 → commit → Part 2 → commit …” sequences with separate commits per part.

Quick Start

Start a rise-pde-session-multi-agents-v2 run by giving your full task description (including any Part 1/Part 2 boundaries) and let it generate the PDE + STC, wait for review, then execute the multi-agent plan while writing the session summary to .pde/<SESSION_UUID>/session-summary.md.