session-memory

What This Skill Does

The session-memory skill acts as the bridge between transient session logs and persistent knowledge storage. It empowers the OpenClaw agent to extract, summarize, and archive past interactions. By default, raw conversation logs are stored as session files, but this skill allows the agent to synthesize these logs into human-readable Markdown files organized by date. Once stored in the designated workspace memory, these files become searchable through standard RAG (Retrieval-Augmented Generation) pipelines, enabling the agent to cite past discussions, recall specific project details, or track the evolution of a conversation over time.

Installation

You can integrate this capability into your OpenClaw environment by running the following command in your terminal: clawhub install openclaw/skills/skills/breezezephyr/session-memory-workspace Ensure that your workspace is correctly configured, as the skill relies on a standard file structure located at ~/.openclaw/workspace for reading sessions and writing daily memory summaries.

Use Cases

• Archiving Daily Progress: Automatically or manually summarize the day's events into a structured memory/YYYY-MM-DD.md file for long-term tracking.
• Knowledge Retrieval: Use the search functionality to scan thousands of conversation lines for specific keywords or topics discussed weeks or months ago.
• Contextual Awareness: Improve the agent's performance by allowing it to 'study' previous interactions before answering questions about ongoing projects or personal preferences.

Example Prompts

1. "总结一下昨天所有的对话，并保存到我的记忆文件中，方便以后查找。"
2. "帮我搜一下上个月我们关于区块链项目的讨论内容，我需要提取一些技术要点。"
3. "把 2026-02-27 的所有聊天记录整理到记忆库里，确保相关的上下文能被以后引用。"

Tips & Limitations

• Formatting: The tool automatically creates a ## Session summary header. Ensure your local memory files maintain this structure to avoid parsing issues.
• Scope: Memory files act as a secondary layer. Use session-search for granular, real-time lookups, and session-to-memory for building a persistent knowledge base.
• Capacity: The default limit for summarization is 200 messages per session. For extremely long technical sessions, you may need to increase the --max-messages parameter to ensure total coverage.
• Consistency: Always verify that your workspace path is correctly set if you are working outside the default ~/.openclaw directory, as the scripts rely on relative path resolution for storage.