Plan-and-Solve Hierarchical Decomposer Skill

High-efficiency, zero-dependency Python implementation of the Plan-and-Solve (PS) prompting and orchestration strategy for complex multi-step reasoning.

Features

• Explicit Hierarchical Planning: Deconstructs convoluted requests into explicit sequential subtasks before execution.
• Dependency Flow: Passes completed task outputs down the pipeline to minimize hallucinations.
• Zero External Dependencies: Pure Python standard library.
• Native MCP Protocol: JSON-RPC 2.0 stdio server compatible with Claude Desktop, Cursor, and Windsurf.

Architecture

graph TD
    Input["Complex Goal / Prompt"] --> Planner["Plan Generation Phase"]
    Planner --> Sub1["Subtask 1: Parse Inputs"]
    Sub1 --> Sub2["Subtask 2: Intermediate Transforms"]
    Sub2 --> Sub3["Subtask 3: Synthesis & Verification"]
    Sub3 --> Output["Final Verified Answer"]