llms-txt

Prompt Decomposition Engine (PDE)

Intention exploration as EAST practice — decomposing prompts into explicit and implicit intents, mapping dependencies, and organizing action through the Four Directions.


PDE as EAST Practice

Prompt Decomposition is part of the EAST direction in the Medicine Wheel framework — the direction of Vision, where:

The EAST practice then guides toward SOUTH — planning and structured activities with human and AI companions. This is the bridge from vision to embodied action.


How PDE Works

A prompt is decomposed into:

Primary Intent

The single most important action — what the prompt is fundamentally asking for.

Secondary Intents

Everything else, distinguished as:

Four Directions Mapping

Direction Role Examples
🌅 East Vision — what is being asked Requirements, desired outcomes, vision statements
🔥 South Analysis — what needs to be learned Research, investigation, reading, growing
🌊 West Validation — what needs reflection Testing, review, accountability checks
❄️ North Action — what executes the cycle Implementation, delivery, creation

Action Stack

An ordered list of tasks respecting dependencies, each mapped to a direction.

Ambiguities

Places where the prompt is vague or uses uncertain language — surfaced explicitly rather than resolved by assumption.


For current agent work, use miaco decompose run as the main operational entrypoint:

miaco decompose run -P prompt.md -e copilot -s iterative-refinement -w .

Use standard when you need the stable baseline or compatibility with existing MCP decomposition flows. Use iterative-refinement when the prompt is layered, recursive, or likely to benefit from multiple readings before action. Treat adversarial-consensus as experimental until the local parser and downstream artifact shape are verified.

The current MCP path (mcp-pde, repo jgwill/mcp-pde) remains the standard decomposition interface. Strategy-aware MCP support is a likely future direction; until it exists, prefer miaco decompose run --strategy ... for strategy selection.


Intent Analyst Role

The Intent Analyst is a PDE-facing role for reading a prompt before execution. Its work is to name the primary creative intent, surface secondary explicit and implicit intents, select the decomposition strategy, and decide whether a parent, child, refinement, or sibling PDE should be created.

The role is especially useful when a prompt contains phrases like “probably,” “maybe,” “I assume,” “what will become,” or references to prior sessions. Those are signals that iterative-refinement may preserve more intent than a single standard pass.


PDE and Delayed Resolution

PDE embodies the delayed resolution principle: ambiguities are named and held rather than silently resolved. This prevents:


Integration

Framework Connection
Creative Orientation PDE frames decomposition as creation, not analysis
Structural Tension Each decomposed intent creates its own tension chart
Narrative Craft Complex prompts contain story — PDE surfaces the narrative structure

Key Sources