memory-consolidation
Periodic "sleep-like" memory optimization. Detects patterns, resolves conflicts, applies decay, and transfers knowledge across contexts.
소스 정보
- 저장소
- kimasplund/clawdbot-skills-pack
- 최근 소스 활동
- 2026년 1월 26일 09:44
- 감지된 SKILL.md 언어
- 영어
- 스타
- 0
- 포크
- 0
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
SKILL.md 표시 중
SKILL.md
소스 지침 · 읽기 전용 미리보기- name
- memory-consolidation
- description
- Periodic "sleep-like" memory optimization. Detects patterns, resolves conflicts, applies decay, and transfers knowledge across contexts.
- metadata
- {"clawdbot":{"requires":{"bins":"[Truncated]"}}}
- user-invocable
- true
# Memory Consolidation
Cross-session learning through periodic memory optimization, inspired by how biological sleep consolidates memories.
## When to Run
| Frequency | Type | Duration | Purpose |
|-----------|------|----------|---------|
| Daily | Light | ~2 min | Health check, quick cleanup |
| Weekly | Standard | ~10 min | Full consolidation |
| Monthly | Deep | ~30 min | System-wide optimization |
## Consolidation Phases
### Phase 1: Pattern Detection
Identify recurring patterns across memories:
- Similar problems solved multiple ways
- Frequently co-accessed memories
- Successful strategies repeated
### Phase 2: Schema Formation
Create higher-level abstractions:
- "When X happens, Y usually works"
- Merge similar memories into schemas
- Reduce redundancy
### Phase 3: Conflict Resolution
Detect and resolve contradictions:
- Same problem, different solutions stored
- Outdated information conflicting with new
- Decide which to keep based on recency + success
### Phase 4: Temporal Decay
Apply time-based forgetting:
- Unused memories decay in Q-value
- Very old, never-accessed memories pruned
- Keep core knowledge stable
### Phase 5: Cross-Context Transfer
Share useful patterns between contexts:
- Debugging insight useful for coding
- Research finding applicable to implementation
- Generalize specific solutions
### Phase 6: QAVR Sync
Update Q-value rankings based on consolidation:
- Boost patterns with high success
- Demote patterns that didn't work
- Rebalance context weights
## Output Report
```markdown
## Memory Consolidation Report
**Date**: [timestamp]
**Type**: [Light/Standard/Deep]
**Duration**: [X minutes]
### Health Metrics
- Total memories: X
- Active (accessed 30d): X
- Dormant (no access 30d): X
- Conflicts detected: X
- Patterns identified: X
### Actions Taken
- Memories merged: X
- Memories pruned: X
- Schemas created: X
- Conflicts resolved: X
- Cross-context transfers: X
### Top Patterns Identified
1. [pattern description]
2. [pattern description]
3. [pattern description]
### Recommendations
- [suggestion for improvement]
### System Health Score
[X/100]
```
## Usage
### Manual Invocation
```
/memory-consolidation daily # Quick health check
/memory-consolidation weekly # Standard consolidation
/memory-consolidation monthly # Deep optimization
```
### Automated Scheduling
Configure in clawdbot.json:
```json
{
"memory": {
"consolidation": {
"daily": "03:00",
"weekly": "sunday 04:00",
"monthly": "1st 05:00"
}
}
}
```
## Benefits
1. **Reduced Noise**: Old, unhelpful memories pruned
2. **Better Retrieval**: Patterns surface useful info
3. **Conflict-Free**: No contradictory advice
4. **Cross-Pollination**: Insights shared across domains
5. **Efficiency**: Less storage, faster queries
GitHub에서 보기