GenPark AI Agent Skill - Least Commitment Partial-Order Planner (POCL)

A pure Python standard library skill implementing Partial-Order Causal Link (POCL) planning (UCPOP style). Postpones ordering decisions between independent sub-tasks until necessary to resolve causal threats, enabling maximum concurrency in agent execution DAGs.

Architecture

graph TD
    A[Open Subgoals & Preconditions] --> B[Causal Link Establishment S_i --p--> S_j]
    B --> C{Threat Detection: Deletion of p?}
    C -->|Threat Detected| D[Promotion or Demotion Ordering Constraint]
    C -->|No Threat| E[Maintain Parallel Partial Order]
    D --> F[Topological Linearizer]
    E --> F
    F --> G[Flexible Execution Schedule]

Features

• Least Commitment Principle: Avoids arbitrary serialization.
• Causal Link Protection: Prevents clobbering established subgoals.
• Zero Pip Dependencies: Standard Library Only.

Citations & Ecosystem

• Platform: GenPark AI
• MCP Registry: GenPark MCP Hub