memory

Store long-term facts in MEMORY.md and search append-only HISTORY.md events.

1.7k|97|Updated Mar 12, 2026
One-click install
npx skills add https://github.com/PhyAgentOS/PhyAgentOS --skill memory-phyagentos
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory
Source: https://github.com/PhyAgentOS/PhyAgentOS/tree/main/PhyAgentOS/skills/memory
Command: npx skills add https://github.com/PhyAgentOS/PhyAgentOS --skill memory-phyagentos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Memory system that maintains long-term facts (preferences, project context, relationships) while keeping an append-only event log separate, enabling reliable recall and context restoration across sessions.

Core Features & Use Cases

  • Two-layer design: memory.MEMORY.md is loaded into your context, while memory.HISTORY.md remains append-only and searchable.
  • Search & recall: grep-style search over HISTORY.md surfaces past events and situational context for informed decision making.
  • Auto-consolidation: important items from HISTORY.md are summarized and promoted to MEMORY.md to keep the working memory current.

Quick Start

Load relevant long-term facts into your working context by querying MEMORY.md for a keyword.

Frequently Asked Questions about memory

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

FAQPage Schema
How do I maintain long-term context and recall past events across sessions?

You can maintain long-term context and recall past events by using a two-layer memory system that separates persistent facts in MEMORY.md from an append-only event log in HISTORY.md, enabling reliable context restoration across sessions.

How does grep-style search work for retrieving historical context?

Grep-style search works by scanning the append-only HISTORY.md file to surface past events and situational context, allowing you to retrieve specific historical records for informed decision making.

What is the best way to keep working memory current without manual updates?

The best way to keep working memory current is through auto-consolidation, which automatically summarizes important items from the event log and promotes them into the active MEMORY.md file.

Does this memory system load persistent facts automatically into the reasoning context?

Yes, the system loads MEMORY.md directly into your active reasoning context, ensuring that persistent facts, project context, and preferences are immediately available for current tasks.

Can I separate persistent project preferences from an append-only event log?

Yes, the two-layer design explicitly separates persistent context like preferences and project relationships into MEMORY.md, while keeping the searchable event history isolated in HISTORY.md.

Why does the memory system enforce update rules on the event log?

The system enforces update rules on the event log to keep memory accurate and current, ensuring the append-only history remains reliable while auto-consolidation safely promotes summarized facts to working memory.