Agent skill

memory-manager

Advanced memory management techniques for the OpenCode memory system

View SKILL.md on GitHub Repository
Stars 0
Forks 0

Install this agent skill to your Project

npx add-skill https://github.com/awaaate/Hyrm/tree/main/.opencode/skill/memory-manager

SKILL.md

Memory Manager Skill

Expert techniques for managing the OpenCode persistent memory system.

Quick Reference

Before any memory operation, consider:

MEMORY ARCHITECTURE

Core Files

File Purpose Format Update Frequency
memory/state.json System state, session count JSON Every session
memory/tasks.json Persistent task tracking JSON On task changes
memory/knowledge-base.json Extracted insights JSON On session end
memory/working.md Current context Markdown Continuous
memory/agent-registry.json Active agents JSON Heartbeat (30s)
memory/message-bus.jsonl Agent messages JSONL On message
memory/sessions.jsonl Session history JSONL On session end

Tools

memory_status()           - Get current state and metrics
memory_search(query)      - Search across memory stores
memory_update(action, data) - Update state/achievements

SMART PRUNING STRATEGIES

Value-Based Retention (What to Keep)

Priority 1 (Always keep):
  - Sessions with code_created > 0
  - Sessions with problems_solved > 0
  - Sessions with quality_score >= 7

Priority 2 (Keep if < 7 days):
  - Sessions with message_count > 20
  - Sessions with key_insights.length > 0

Priority 3 (Prune after 3 days):
  - Heartbeat-only sessions
  - Sessions with error status

Pruning Commands

bash
# Analyze what would be pruned
bun tools/knowledge-deduplicator.ts analyze

# Archive old sessions (> 7 days)
bun tools/working-memory-manager.ts archive

# Clean message bus (remove old heartbeats)
bun tools/message-bus-manager.ts cleanup --days 3

Message Bus Cleanup

bash
# Keep:
- task_complete messages (30 days)
- request_help messages (7 days)  
- broadcast announcements (7 days)

# Remove:
- Heartbeat messages (> 24 hours)
- Read direct messages (> 7 days)

KNOWLEDGE EXTRACTION

Extraction Process

bash
# Extract from current session
bun tools/knowledge-extractor.ts extract

# Extract from specific session
bun tools/knowledge-extractor.ts extract --session <session-id>

# Rebuild entire knowledge base
bun tools/knowledge-extractor.ts rebuild

What Gets Extracted

- Key insights and learnings
- Decisions made and rationale
- Problems solved and solutions
- Code patterns discovered
- Tool usage patterns

MEMORY OPTIMIZATION

Deduplication

bash
# Analyze duplicates in knowledge base
bun tools/knowledge-deduplicator.ts analyze

# Remove duplicates (dry run first!)
bun tools/knowledge-deduplicator.ts dedupe --dry-run
bun tools/knowledge-deduplicator.ts dedupe

Token Management

Thresholds:
- Normal: < 100k tokens
- Warning: 100k-150k tokens (suggest pruning)
- Critical: > 150k tokens (require pruning)

Actions:
- Check with: memory_status(include_metrics=true)
- Archive old working memory: bun tools/working-memory-manager.ts archive
- Prune knowledge base: bun tools/knowledge-deduplicator.ts dedupe

VALIDATION

JSON Validation

bash
# Validate memory files
bun -e "JSON.parse(require('fs').readFileSync('memory/state.json'))"
bun -e "JSON.parse(require('fs').readFileSync('memory/tasks.json'))"
bun -e "JSON.parse(require('fs').readFileSync('memory/knowledge-base.json'))"

Integrity Checks

bash
# Full health check
bun tools/cli.ts memory health

# Check for orphaned tasks
bun tools/task-manager.ts orphans

TROUBLESHOOTING

Common Issues

Corrupted JSON

bash
# Check syntax errors
bun -e "JSON.parse(require('fs').readFileSync('memory/state.json'))" 2>&1

# Restore from backup
cp memory/backups/state-latest.json memory/state.json

Race Conditions in Multi-Agent

  • Use task_claim for atomic operations
  • Check file locks before major updates
  • Monitor message-bus for conflicts

Growing Memory

bash
# Check file sizes
ls -lh memory/*.json

# Archive old data
bun tools/working-memory-manager.ts archive

BEST PRACTICES

Regular Maintenance Schedule

Daily:
  - Check memory_status() for issues
  - Review quality_report() trends

Weekly:
  - Run knowledge deduplication
  - Archive old working memory
  - Clean message bus

Monthly:
  - Full backup of all memory files
  - Review and prune knowledge base
  - Performance analysis

Backup Strategy

bash
# Create dated backup
mkdir -p memory/backups
cp memory/state.json memory/backups/state-$(date +%Y%m%d).json
cp memory/knowledge-base.json memory/backups/kb-$(date +%Y%m%d).json

Remember: The memory system is the brain of the AI. Keep it healthy!

Recommended Agent Skills

Expand your agent's capabilities with these related and highly-rated skills.

awaaate/Hyrm

agent-orchestrator

Orchestrator agent for coordinating multiple AI agents in a persistent system

0 0
Explore
awaaate/Hyrm

critique-agent

Specialized agent for code review, debugging critique, and quality assessment

0 0
Explore
awaaate/Hyrm

multi-agent-patterns

Common patterns and best practices for multi-agent coordination

0 0
Explore
mattpocock/skills

handoff

Compact the current conversation into a handoff document for another agent to pick up.

111,310 9,758
Explore
mattpocock/skills

setup-pre-commit

Set up Husky pre-commit hooks with lint-staged (Prettier), type checking, and tests in the current repo. Use when user wants to add pre-commit hooks, set up Husky, configure lint-staged, or add commit-time formatting/typechecking/testing.

111,310 9,758
Explore
mattpocock/skills

obsidian-vault

Search, create, and manage notes in the Obsidian vault with wikilinks and index notes. Use when user wants to find, create, or organize notes in Obsidian.

111,310 9,758
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results