agent-memory

name	agent-memory
description	Persistent memory system for AI agents with semantic search and context management

Agent Memory

This built-in skill provides a persistent memory system for AI agents to store, retrieve, and manage contextual information across sessions and interactions.

Capabilities

• Memory Storage: Store structured and unstructured data with metadata
• Semantic Search: Retrieve memories using natural language queries with vector similarity
• Context Management: Maintain conversation context and relevant memories
• Memory Compression: Compress and summarize long-term memories for efficiency
• Temporal Awareness: Track when memories were created and accessed
• Memory Linking: Create relationships between related memories
• Privacy Controls: Manage memory access permissions and data retention
• Export/Import: Export memories to various formats and import from external sources
• Memory Cleanup: Automatically clean up outdated or irrelevant memories
• Multi-modal Support: Store and retrieve text, images, and other data types

Usage Examples

Store Memory

tool: agent-memory
action: store
memory:
  content: "User prefers dark mode and uses VS Code as their primary editor"
  tags: ["user_preferences", "development"]
  context: "conversation_12345"
  importance: 0.8

Retrieve Memory

tool: agent-memory
action: retrieve
query: "What are the user's development preferences?"
context: "current_conversation"
limit: 5

Update Memory

tool: agent-memory
action: update
memory_id: "mem_67890"
updates:
  content: "User now prefers light mode during daytime hours"
  tags: ["user_preferences", "development", "ui_theme"]

Security Considerations

• All memory data is encrypted at rest using industry-standard encryption
• Access control ensures only authorized agents can access specific memories
• Data retention policies automatically expire sensitive information
• Privacy-by-design principles prevent unauthorized data collection
• Audit logging tracks all memory operations for compliance

Configuration

The agent-memory skill can be configured with the following parameters:

• storage_backend: Storage backend (sqlite, postgres, memory)
• embedding_model: Embedding model for semantic search (default: sentence-transformers)
• max_memory_size: Maximum size of individual memories (default: 10KB)
• retention_policy: Data retention policy (default: 30 days for non-essential data)
• privacy_level: Privacy level controls (strict, moderate, relaxed)
• compression_enabled: Enable memory compression (default: true)

This skill is essential for any agent that needs to maintain context across interactions, learn from past experiences, or provide personalized responses based on historical data.