lucy-workspace-openclaw

Summary

This workspace organizes files and tools for developing, tracking, and improving self-improving agent skills within OpenClaw environments. It provides structure for agent identity, memory management, local tool configurations, and a complete workflow for capturing and applying learnings to enhance agent performance over time.

Included Assets

• AGENTS.md: Guidance for session startup, agent identity, and memory handling.
• SOUL.md: Defines agent personality, core truths, and operational boundaries.
• MEMORY.md: Long-term curated memory for major events, lessons, and preferences.
• TOOLS.md: Local configuration notes for devices, SSH, and other environment-specific tools.
• memory/YYYY-MM-DD.md: Per-day session notes.
• skills/self-improving-agent/
  • .learnings/ERRORS.md: Log of errors and failures.
  • .learnings/FEATURE_REQUESTS.md: User-requested features not yet available.
  • .learnings/LEARNINGS.md: Corrections, insights, and knowledge gaps with status tracking.
  • SKILL.md: Documentation of the self-improvement skill, detailed triggers, and logging guidance.
  • assets/LEARNINGS.md: Category/status reference and sample formatting for learnings.
  • assets/SKILL-TEMPLATE.md: Templates for extracting new skills from learnings.
  • hooks/openclaw/HOOK.md, handler.js, handler.ts: Bootstrap hook to remind agents to capture learnings at startup.
  • references/examples.md: Real-world entry examples for structured learnings.
  • references/hooks-setup.md: Guide for configuring automatic self-improvement reminders.

How to Use

1. Session Initialization:

   • Read SOUL.md for agent identity and boundaries.
   • Use AGENTS.md for session start process and memory management.
   • Review the latest memory/YYYY-MM-DD.md for recent context.
   • Update MEMORY.md in direct sessions for curated, long-term memory.

2. Tooling and Local Setup:

   • Record environment-specific info (e.g., device names, SSH hosts) in TOOLS.md.
   • Respect restricted folders as specified in TOOLS.md.

3. Self-Improvement Workflow:

   • Capture all errors in .learnings/ERRORS.md.
   • Record user corrections and new insights in .learnings/LEARNINGS.md.
   • Log missing features in .learnings/FEATURE_REQUESTS.md.
   • Reference assets/LEARNINGS.md and assets/SKILL-TEMPLATE.md for consistent formatting.
   • Use hooks/openclaw to trigger learning reminders at agent bootstrap.
   • Promote important learnings to skills using the provided templates.

4. Extending and Sharing:

   • Review SKILL.md and reference material for developing or extracting additional agent skills.
   • Use examples and guides from the references/ directory to maintain consistent and actionable logs.

Notes

• Always write down key events or learnings—memory files should reflect important developments.
• Long-term memory (MEMORY.md) is only to be accessed in trusted, direct user-agent sessions for privacy.
• Never access or log sensitive folders/items listed as restricted in TOOLS.md.
• Update, promote, or extract learnings into reusable skills to foster continuous improvement.
• All logging and improvement processes are openly documented within the workspace for transparency and reproducibility.