/memory

name: diverga-memory
description: |
  Diverga Memory System v7.0 - Context-persistent research support
  with checkpoint auto-trigger and cross-session continuity.
  Triggers: memory, remember, context, recall, checkpoint, decision, persist,
  기억, 맥락, 세션, 체크포인트
version: "12.0.1"

Diverga Memory System v7.0

Overview

Human-centered research context persistence with:

• 3-Layer Context System
• Checkpoint Auto-Trigger
• Cross-Session Continuity
• Decision Audit Trail
• Research Documentation Automation

Quick Reference

Context Loading Keywords

**English**: "my research", "research status", "where was I", "continue research", "what stage"

**Korean**: "내 연구", "연구 진행", "연구 상태", "어디까지", "지금 단계"

Commands

| Command | Description | |---------|-------------| | `/diverga:memory status` | Show project status | | `/diverga:memory context` | Display full context | | `/diverga:memory init` | Initialize project | | `/diverga:memory decision list` | List decisions | | `/diverga:memory archive [STAGE]` | Archive stage | | `/diverga:memory migrate` | Run migration |

Priority Context (v8.2 — Compression Resilience)

MCP Tools for Priority Context

| Command | MCP Tool | Description | |---------|----------|-------------| | Read priority | `diverga_priority_read()` | Read 500-char context summary | | Write priority | `diverga_priority_write(context)` | Update context summary | | Full status | `diverga_project_status()` | Project state + checkpoints + decisions | | Check prereqs | `diverga_check_prerequisites(agent_id)` | Verify agent can proceed | | Record decision | `diverga_mark_checkpoint(cp_id, decision, rationale)` | Record and auto-update priority |

Auto-Update Behavior

Priority context is automatically updated when:

• A checkpoint is marked via `diverga_mark_checkpoint()`
• Format: `Project: {name} | Paradigm: {paradigm} | RQ: {question} | ✅/❌ checkpoints | Last: {decision}`
• Maximum 500 characters, stored at `.research/priority-context.md`

Compression Recovery

When context window is compressed: 1. Call `diverga_priority_read()` to recover essential project context 2. Call `diverga_checkpoint_status()` to see checkpoint state 3. Call `diverga_project_status()` for full project details

3-Layer Context System

Layer 1: Keyword-Triggered (자연어 감지)

When researcher asks "내 연구 진행 상황은?" or "What's my research status?", automatically load and display context.

**Auto-Detection Keywords**:

• "my research", "연구", "research", "progress", "진행"
• "where was I", "continue", "다시", "어디까지"
• "what stage", "현재 단계", "stage", "지금"

**Response Pattern**: 1. Detect keyword match 2. Load `.research/project-state.yaml` 3. Display current stage and progress 4. Show pending checkpoints 5. List available next actions

Layer 2: Task Interceptor (에이전트 호출)

When `Task(subagent_type="diverga:*")` is called, automatically inject full research context and checkpoint instructions.

**Injection Process**: 1. Detect `diverga:` prefix in subagent_type 2. Read `.research/project-state.yaml` 3. Read `.research/checkpoints.yaml` 4. Inject context into agent prompt 5. Add checkpoint validation wrapper 6. Execute with full research awareness

**Context Injected**:

# Automatically included in agent prompt
research_context:
  project_name: "[from project-state.yaml]"
  current_stage: "[from checkpoints.yaml]"
  research_question: "[from project-state.yaml]"
  methodology: "[from project-state.yaml]"
  decisions: "[from decision-log.yaml, last 10]"
  pending_checkpoints: "[from checkpoints.yaml]"

Layer 3: CLI (명시적 요청)

Run `/diverga:memory context --verbose` for full detailed state.

**Available Flags**:

• `--verbose` - Show full decision audit trail
• `--archive` - Include archived stages
• `--decisions` - Show decision log only
• `--checkpoints` - Show checkpoint status only
• `--format json|yaml|text` - Output format

Checkpoint System

Checkpoint Levels

| Level | Icon | Behavior | Example | |-------|------|----------|---------| | REQUIRED | 🔴 | Must complete before proceeding | CP_RESEARCH_DIRECTION | | RECOMMENDED | 🟠 | Strongly suggested | CP_PARADIGM_SELECTION | | OPTIONAL | 🟡 | Can skip with defaults | CP_METHODOLOGY_APPROVAL |

Standard Checkpoints (Research Workflow)

Foundation Stage (0-2 hours)

• **CP_RESEARCH_DIRECTION** 🔴 - Research question finalized and validated
• **CP_PARADIGM_SELECTION** 🟠 - Quantitative/qualitative/mixed selected with rationale
• **CP_SCOPE_DEFINITION** 🔴 - Scope constraints documented (years, populations, outcomes)

Design Stage (2-4 hours)

• **CP_THEORY_SELECTION** 🟠 - Theoretical framework chosen and justified
• **CP_VARIABLE_DEFINITION** 🔴 - All variables operationalized (IV, DV, mediators, moderators)
• **CP_METHODOLOGY_APPROVAL** 🟠 - Research design validated (RCT, meta-analysis, qualitative, etc.)

Planning Stage (4-6 hours)

• **CP_DATABASE_SELECTION** 🔴 - Data sources identified with inclusion/exclusion criteria
• **CP_SEARCH_STRATEGY** 🔴 - Search terms, filters, and retrieval approach documented
• **CP_SAMPLE_PLANNING** 🟠 - Sample size, power analysis (if quantitative), or saturation plan (if qualitative)

Execution Stage (6+ hours)

• **CP_SCREENING_CRITERIA** 🔴 - Inclusion/exclusion criteria operationalized for systematic review
• **CP_RAG_READINESS** 🟠 - Vector database and retrieval system configured
• **CP_DATA_EXTRACTION** 🟠 - Data extraction protocol finalized and tested
• **CP_ANALYSIS_PLAN** 🔴 - Analysis approach documented with reproducible steps

Validation Stage (Final)

• **CP_QUALITY_GATES** 🔴 - PRISMA/CONSORT compliance verified
• **CP_PEER_REVIEW** 🟠 - Methodology reviewed by co-investigators
• **CP_PUBLICATION_READY** 🔴 - Manuscript format and ethics approved

Checkpoint Enforcement Rules

**REQUIRED (🔴) Checkpoints**:

• Cannot skip
• Must have evidence of completion
• Blocks advancement to next stage
• Tracked in `decision-log.yaml` with timestamp