Back to Interpretive Loop Diagrams

Prototype 03

ooda_loop.py

SPINE OODA Loop

Observe, Orient, Decide, Act — the core reasoning cycle for Adaptivearts.ai agent workflows. Each cycle passes through 6-layer context stacks with MemoryHooks integration for persistent decision tracking.

OODA Phases

Four-phase reasoning cycle with context-aware orientation

1

Observe

Gather inputs from context stacks, tool outputs, and environment state. Raw signal collection.

2

Orient

Interpret observations through 6-layer context stacks. Apply constraints, character context, and prior decisions.

3

Decide

Select next action from available API tools based on oriented understanding. MemoryHooks integration for decision history.

4

Act

Execute the chosen tool call, capture output, and feed results back into the next Observe phase.

6-Layer Context Stacks

Structured context shaping the Orient phase

Global

System-wide instructions and behavioral constraints

Character

Agent persona, tone, and domain expertise

Command

Current task directive and success criteria

Constraints

Guardrails, safety rules, and output format requirements

Context

Session history, prior outputs, and accumulated knowledge

Input

Current user or agent input triggering this cycle

Integration Points

Connects context engineering to action execution

MemoryHooks

Persists decisions, observations, and outcomes across OODA cycles.

context-glue-mcp

Budget-managed context assembly for the 6-layer stack composition.

Additional Materials

Study materials coming soon.