cross-session-memory

What This Skill Does

The cross-session-memory skill acts as the long-term cognitive engine for the OpenClaw AI agent. It solves the inherent problem of AI statelessness by implementing a rolling 24-hour memory retrieval system combined with long-term curated memory files. By automatically parsing daily log files (memory/YYYY-MM-DD.md) and integrating your core personal preferences from MEMORY.md, it ensures the agent maintains context across days, sessions, and project phases. It essentially creates a seamless, ongoing conversation that doesn't "reset" when the terminal restarts, allowing for high-level continuity in complex workflows.

Installation

To install this skill, use the clawhub command: clawhub install openclaw/skills/skills/2426758093/cross-session-memory Once installed, integrate it by adding the script path to your OpenClaw sessionStart hook in your agent configuration, or follow the instructions in AGENTS.md to ensure the agent reads the daily memory files as a prerequisite task upon initialization.

Use Cases

• Project Continuity: Working on a code base over several days and needing the agent to remember technical constraints, API keys, or current tasks from yesterday.
• Persona Consistency: Ensuring the agent remembers user preferences (e.g., tone of voice, preferred formatting) across different sessions.
• Task Tracking: Keeping a history of completed versus failed tasks, preventing the agent from re-proposing solutions that have already been identified as non-viable.
• Remote Collaboration: Enabling the agent to pick up exactly where you left off after an overnight break, reducing the need to repeat background information.

Example Prompts

1. "Based on our progress yesterday, what was the last task we blocked on and how should we proceed today?"
2. "Review the notes from the last 3 days in our memory files and summarize the key technical decisions we made regarding the news automation system."
3. "Check my personal preferences in MEMORY.md and ensure the report you generate for me strictly follows those guidelines."

Tips & Limitations

• Curate Your Memory: While the agent handles daily raw logs automatically, the MEMORY.md file requires manual updates to keep your long-term goals accurate.
• File Management: Keep your memory/ folder clean; the script uses Get-ChildItem to sort by date, so ensure filenames follow the YYYY-MM-DD.md format strictly.
• Performance: For massive project histories, stick to the RollingHours parameter to keep response times fast. Loading several months of logs into the context window can increase token consumption significantly.