openclaw-workspace

Summary

This workspace is a template for building, managing, and extending AI agents in OpenClaw. It provides structure and tools for agent configuration, memory management, self-improvement, and GitHub integration. The layout supports session continuity, secure and clear memory practices, and reusable skills for automation and proactivity.

Included Assets

• AGENTS.md: Session and workspace operating instructions for agents
• SOUL.md: Agent guiding principles, boundaries, and personality notes
• TOOLS.md: Overview of installed skills, available tools, and practical commands
• skills/ai/github-cli/: GitHub CLI skill with detailed references and metadata
• skills/memory/self-improving/: Self-improving agent skill, with guidance for correction logging, boundary rules, memory, heartbeat maintenance, and learning mechanics

How to Use

1. Agent Session Startup:

   • Read SOUL.md to understand your role and personality model.
   • Follow the instructions in AGENTS.md for session startup and context restoration:
     • Load recent memory files for session continuity (memory/YYYY-MM-DD.md, MEMORY.md).
     • Operate according to the outlined routines and session context.

2. Skills and Tools:

   • Review TOOLS.md to see all installed skills, planned installations, system tools, and example commands for OpenClaw operation and skill management.
   • Skills are organized under skills/ by purpose (e.g., AI/GitHub integration and self-improvement).
   • To install new skills or check status, use commands from the "常用命令" section in TOOLS.md.

3. Memory and Self-Improvement:

   • The self-improving skill stores and updates agent memory, correction logs, and preferences in ~/.openclaw/self-improving/.
   • Adjust and review boundaries, corrections, and learning behavior via the files in skills/memory/self-improving/.
   • Follow the heartbeat rules to keep self-improvement data properly maintained.

4. GitHub Integration:

   • For GitHub automation, use the github-cli skill (requires the gh command-line tool).
   • Configuration and reference are documented in skills/ai/github-cli/SKILL.md.

Notes

• Sensitive data must never be stored in agent memory (see boundaries.md for specifics).
• Memory is session-based for privacy and safety; continuity relies on the proper use and upkeep of designated memory files.
• The workspace supports both English and Chinese tool documentation.
• The heart of agent improvement is self-critique and user correction. Review and update your memory and correction logs regularly.
• For initial setup, certain files (as described in templates) are created or should be referenced under the user’s OpenClaw directory (~/.openclaw/self-improving/).

For more detail, see the markdown files under each path, especially AGENTS.md, SOUL.md, TOOLS.md, and the respective SKILL.md files for each skill.