Intention exploration as EAST practice — decomposing prompts into explicit and implicit intents, mapping dependencies, and organizing action through the Four Directions.
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.
A prompt is decomposed into:
The single most important action — what the prompt is fundamentally asking for.
Everything else, distinguished as:
| 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 |
An ordered list of tasks respecting dependencies, each mapped to a direction.
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.
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 embodies the delayed resolution principle: ambiguities are named and held rather than silently resolved. This prevents:
| 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 |