rise-pde-session-multi-agents-v2

Multi-agent orchestration for multi-file software changes is handled by decomposing user intent into a structured PDE action stack, then coordinating parallel deep-search, validation, and execution agents with strict file-ownership boundaries. It generates an STC chart for human review before execution proceeds.

PDE decomposition is a structured workflow that breaks a user's intent into an actionable stack, generating a structural tension chart for review. It supports cascading multi-part missions by detecting sequential boundaries and executing them with separate commits per part.

Yes, you can review and edit the action plan before execution. The workflow enforces a bash sleep 1500 pause, giving humans time to review and modify the authoritative .pde/ files before NORTH executor agents resume the coordinated implementation.

Kinship-aware documentation and rispecs are managed during repository updates by deploying cross-reference validator and executor agents that maintain structural awareness. These agents operate with file-ownership boundaries to ensure documentation aligns with the multi-file changes.

Yes, multi-agent PDE session execution isolates workspace files for each run by strictly enforcing a dedicated workdir under .pde/<SESSION_UUID>/. This ensures all spawned subagents, parameter constraints, and session summaries remain contained within that specific session directory.

The best way to handle cascading multi-part missions is using a PDE workflow that detects sequential Part 1 to Part 2 boundaries and executes them with separate commits per part. The multi-agent plan coordinates research, validation, and implementation across each committed segment.