memory

Manage long-term facts and short-term context with MEMORY.md and HISTORY.md files.

55|7|Updated Mar 2, 2026
One-click install
npx skills add https://github.com/wp931120/tiny_agent --skill memory-wp931120
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory
Source: https://github.com/wp931120/tiny_agent/tree/main/workspace/skills/memory
Command: npx skills add https://github.com/wp931120/tiny_agent --skill memory-wp931120

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Two-layer memory system with grep-based recall to persist important facts and provide quick access to past context.

Core Features & Use Cases

  • Two-layer memory: MEMORY.md for long-term facts and HISTORY.md for append-only logs.
  • Auto-consolidation: old conversations summarized and merged to long-term memory.
  • Quick retrieval: grep-based search to locate past events and facts when needed.

Quick Start

Store a key fact in MEMORY.md to enable future recall in conversations.

Frequently Asked Questions about memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I implement long-term memory for AI agents to recall past context?

Long-term memory for AI agents is managed using a two-layer architecture with MEMORY.md for persistent facts and HISTORY.md for append-only logs. This setup enables persistent recall across interactive sessions and project contexts.

How does auto-consolidation work for AI conversation history?

Auto-consolidation works by summarizing old conversations and merging them into long-term memory. This process transfers relevant past context from short-term history logs into persistent facts for continuous future use.

What is the best way to grep past events and facts for AI context management?

The best way to grep past events and facts is through grep-based retrieval on a two-layer memory system. This allows quick search across append-only logs and consolidated long-term facts to locate required context.

Can I use this two-layer memory architecture across multiple project contexts?

Yes, the two-layer memory architecture applies across multiple project contexts. It organizes memory to support memory-driven workflows requiring recall of past information within various interactive sessions.

Why do I need separate files for short-term context and long-term facts?

Separating short-term context and long-term facts optimizes context management by isolating append-only logs from persistent facts. This prevents history bloat while ensuring quick retrieval of consolidated, relevant information.