memory-archiver

What This Skill Does

The Memory Archiver is an advanced knowledge management engine designed for OpenClaw. It implements a multi-layer temporal storage architecture (hourly to yearly) to ensure that critical context is preserved, summarized, and archived. By leveraging automated cron tasks, the agent intelligently filters noise from daily interactions, capturing vital decisions, technical configurations, and problem-solving milestones while maintaining a clean, searchable file structure.

Installation

To integrate the skill, execute the following command in your terminal: clawhub install openclaw/skills/skills/amd5/memory-archiver Alternatively, for manual setup, place the script files in your ~/.openclaw/workspace/skills/memory-manager directory and run the registration commands provided in the official documentation to activate the cron-based scheduling triggers.

Use Cases

• Project Long-Term Continuity: Retain technical context across months, ensuring the agent remembers architectural decisions made weeks ago.
• Automated Logging: Maintain a consistent, chronological journal of agent activities and configuration changes without manual input.
• Refinement Cycles: Compress high-frequency hourly logs into concise daily summaries and eventually into monthly/yearly retrospectives for high-level oversight.

Example Prompts

1. "OpenClaw, please summarize the key technical blockers we solved yesterday by pulling from the memory-archiver."
2. "What was the agreed-upon configuration for the database migration that we decided on last month?"
3. "Show me a summary of our progress from this past week based on the weekly archival file."

Tips & Limitations

• Tips: Use the memory/daily/ directory as your primary source for quick factual retrieval; use the weekly/ summaries for strategic planning.
• Limitations: This skill assumes file-write permissions are available. Ensure your system environment has sufficient disk space for long-term storage of text-based archives. Avoid manual edits to the hourly files while a scheduled write-task is running to prevent race conditions. The archive relies heavily on the quality of the 'important information' identification filter; if the agent is too broad, you may need to adjust the filtering prompt in the scripts/sync-daily.sh file.