memory

Retrieve persistent conversation memory via grep-style search from MEMORY.md and HISTORY.md.

6|2|Updated Apr 4, 2026

Search Tags:#context #recall #search #memory #persistence #history #grep

Ready to run

npx skills add https://github.com/zeminroot/cheerclaw --skill memory-zeminroot

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: memory
Source: https://github.com/zeminroot/cheerclaw/tree/main/cheerclaw/skills/memory
Command: npx skills add https://github.com/zeminroot/cheerclaw --skill memory-zeminroot

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

双层记忆系统，基于 grep 检索召回记忆信息。它提供跨会话的持久记忆能力，帮助保持上下文、偏好与关键事实在长期对话中的可用性。

Core Features & Use Cases

双层结构：origin_qas.jsonl 保存原始对话日志，HISTORY.md 保存追加的事件摘要，MEMORY.md 保存长期重要事实，确保核心信息始终可加载。
快速检索：使用 grep 风格搜索在历史记忆中定位相关条目，并在需要时加载到上下文中进行筛选。
典型用例：适用于跨会话记忆、用户偏好管理、长期项目跟踪和知识累积等场景。

Quick Start

Ask the agent to retrieve memory items related to a topic from MEMORY.md and HISTORY.md using grep-style search.

Frequently Asked Questions about memory

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

FAQPage Schema

How do I maintain persistent context across agent conversations?▼

Persistent context across agent conversations is maintained using a two-layer memory system that stores original logs in origin_qas.jsonl, event summaries in HISTORY.md, and key facts in MEMORY.md for continuous recall.

How can I search historical agent memory using grep-style retrieval?▼

Historical agent memory is searchable via grep-style retrieval to locate relevant entries in MEMORY.md and HISTORY.md, loading only matched items into the context for precise filtering and validation.

What is the best way to manage user preferences in long-running support tasks?▼

Managing user preferences in long-running support tasks is best handled by incrementally updating a persistent MEMORY.md file, ensuring core preference facts remain available across sessions.

Does this cross-session memory approach store original conversation logs?▼

This cross-session memory approach stores original conversation logs in the origin_qas.jsonl file, separating raw data from summarized events in HISTORY.md and long-term facts in MEMORY.md.

When do I need a two-layer memory model for agent context?▼

A two-layer memory model for agent context is needed during onboarding and long-running support tasks where preserving historical logs and summarizing incremental events ensures continuous knowledge accumulation.

Are there limitations to loading full conversation history into the agent context?▼

Loading full conversation history into the agent context risks exceeding token limits, so controlled loading via grep-style search retrieves only relevant memory items to maintain robust validation.