memory

Maintain long-term facts in MEMORY.md and grep-searchable events in HISTORY.md.

11|Updated Feb 28, 2026
One-click install
npx skills add https://github.com/PrettiFlow/pretticlaw --skill memory-prettiflow
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory
Source: https://github.com/PrettiFlow/pretticlaw/tree/main/src/skills/memory
Command: npx skills add https://github.com/PrettiFlow/pretticlaw --skill memory-prettiflow

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Provides a persistent, two-layer memory so agents can retain long-term facts and efficiently search past events without overloading session context.

Core Features & Use Cases

  • Always-loaded long-term facts: memory/MEMORY.md stores user preferences, project context, and relationship facts that are kept in context.
  • Grep-searchable event log: memory/HISTORY.md is an append-only event history designed for fast, deterministic grep queries and is not loaded into context.
  • Auto-consolidation: older conversations are summarized into HISTORY.md with extracted facts appended to MEMORY.md, reducing manual maintenance.
  • Use Case: During a development session, the agent reads MEMORY.md for configuration preferences and performs a grep on HISTORY.md to find past meeting notes or deadlines.

Quick Start

Ask the agent to recall the user's preference from MEMORY.md and search HISTORY.md for the keyword meeting using a case-insensitive grep.

Frequently Asked Questions about memory

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

FAQPage Schema
How do I add long-term memory and recall past events for a conversational agent?

You can search past events in agent history using exec-based deterministic shell grep queries on the append-only HISTORY.md file, which is specifically designed for fast and precise keyword retrieval.

Can I store user preferences and project context without overloading session context?

You can store user preferences and project context in MEMORY.md, which remains always loaded, while past events are auto-consolidated into HISTORY.md to prevent overloading the active session context.

How does auto-consolidation extract facts from older conversations?

Auto-consolidation automatically summarizes older conversations into the HISTORY.md event log and appends extracted facts to MEMORY.md, which reduces manual maintenance and preserves long-term memory accuracy.

Does this two-layer memory approach work for developer workflows needing project context?

This two-layer memory approach works for developer workflows by keeping configuration preferences in MEMORY.md and allowing agents to grep HISTORY.md for specific past meeting notes or deadlines during sessions.

What are the limitations of using shell grep to search an append-only event history?

A limitation of using shell grep to search an append-only event history is that HISTORY.md is not loaded into context, meaning you must perform explicit exec-based deterministic searches rather than relying on automatic semantic recall.