memory-coordination

# Memory Coordination

## Overview

**Purpose**: Memory Coordination is Shannon's protocol for structured Serena MCP operations. It enforces standardized entity naming (shannon/* namespace), mandatory relation creation, consistent search patterns, and proper observation management to maintain a clean, queryable knowledge graph across all waves and sessions.

**When to Use**:
- Storing ANY Shannon artifact (spec, wave, goal, checkpoint, SITREP)
- Querying Shannon history (waves, checkpoints, specs)
- Creating entities for Shannon context preservation
- Updating existing entities with new information
- Establishing relations between Shannon entities
- Searching Shannon knowledge graph

**Expected Outcomes**:
- Standardized entity names with shannon/* namespace
- Complete relational graph (no orphaned entities)
- Efficient queries using consistent patterns
- Cross-wave context preservation
- Zero data duplication via observations
- Clean knowledge graph structure

**Duration**: 10-30 seconds per operation

---

## Anti-Rationalization (From Baseline Testing)

**CRITICAL**: Agents systematically rationalize skipping memory-coordination protocols. Below are the 6 most common rationalizations detected in baseline testing, with mandatory counters.

### Rationalization 1: "Entity name seems fine"
**Example**: Agent creates entity named "spec_data" or "MyAnalysis" without shannon/ prefix

**COUNTER**:
- ❌ **NEVER** create Shannon entities without shannon/* namespace
- ✅ ALL Shannon entities MUST start with shannon/
- ✅ Use standard namespaces: shannon/specs/, shannon/waves/, shannon/goals/, shannon/checkpoints/, shannon/sitreps/
- ✅ If name lacks shannon/ prefix, it's WRONG

**Rule**: All Shannon entities have shannon/* namespace. No exceptions.

### Rationalization 2: "Relations not needed for this entity"
**Example**: Agent creates wave entity without relating it to spec

**COUNTER**:
- ❌ **NEVER** create standalone Shannon entities
- ✅ EVERY entity has at least 1 relation