openclaw-workspace-longpan

Summary

This workspace provides structure, templates, and guides for building, improving, and managing self-improving agents within OpenClaw. It includes agent identity files, memory logs, customizable environment notes, and a skillset dedicated to continuous improvement and error tracking. Core concepts emphasize clear memory management, agent personality, and actionable learning.

Included Assets

• AGENTS.md — Workspace instructions, session routines, memory management workflow
• MEMORY.md — Long-term memory; identity, timeline, important decisions (中文)
• SOUL.md — Agent personality, values, and behavioral guidelines
• TOOLS.md — Environment-specific notes (devices, aliases, room names, etc.)
• memory/2026-03-18.md — Example daily memory log
• skills/self-improving-agent/ — Self-improvement skillset and resources:
  • .clawhub/origin.json, _meta.json — Skill metadata
  • .learnings/ERRORS.md, FEATURE_REQUESTS.md, LEARNINGS.md — Logs for errors, features, learnings
  • SKILL.md — Skill documentation (when to log issues, how to extract learnings)
  • assets/LEARNINGS.md, SKILL-TEMPLATE.md — Templates for learning/skill extraction and documentation
  • hooks/openclaw/HOOK.md, handler.js/handler.ts — Hook to inject reminders for self-improvement during agent bootstrap
  • references/examples.md — Example entries for logs
  • references/hooks-setup.md — Guide for automating learning capture via hooks

How to Use

1. Set Up Agent Memory:

   • Daily use: Create memory/YYYY-MM-DD.md logs for notable events and session context.
   • Curate valuable insights in MEMORY.md for long-term reference (only in private sessions).

2. Define Agent Personality:

   • Use SOUL.md to set behavioral boundaries, values, and assistant philosophy.
   • Update as personality evolves; inform the user about major changes.

3. Customize Environment:

   • Document local tools/devices in TOOLS.md for reference by agent and human.

4. Capture Learnings and Improvements:

   • Log failures, corrections, and feature requests in the .learnings/ markdown files.
   • Review learnings before tackling major tasks or upgrades.
   • Use asset templates to extract repeatable skills or workflow improvements.

5. Automate Self-Improvement Reminders:

   • Enable self-improvement hooks via workspace or skill files; reminders prompt logging after agent bootstrap.
   • See references/hooks-setup.md for integrating with agent scripts or Claude settings.

6. Skill Management:

   • Install skills as described in relevant files; use SKILL.md for action guides.
   • Promote critical learnings to memory or agent core files as appropriate.

Notes

• Long-term memory (MEMORY.md) should only be accessed during main (private) sessions for privacy reasons.
• All logs and templates are markdown-based for transparency and ease of editing.
• The workspace encourages explicit, written memory — no "mental notes," always write down what matters.
• Examples and templates are provided for consistent documentation of learnings and skill extractions.
• Asset files and logs can be extended or reorganized as agent needs and complexity grow.
• The agent identity and behavior guidelines evolve with use; major changes should be communicated to the user.
• Environment notes (TOOLS.md) are kept separate from reusable skills to simplify upgrades and sharing.