jarvis-workspace

Summary

This workspace provides a comprehensive setup for building and managing OpenClaw-based AI agents. It includes agent configuration files, skills, persistent memory logs, and documentation to support multi-agent collaboration, skill management, and long-term context retention. The main focus is on practical workflow, agent identity, user preferences, and technical notes, making it suitable for real-world deployment and daily operation in team environments.

Included Assets

• Agent & Identity

  • AGENTS.md — Agent role, session rules, etiquette, and operational conventions
  • SOUL.md — Agent core identity, communication principles, and boundaries
  • TOOLS.md — Local tool notes, workspace structure, multi-agent architecture

• Memory & Context

  • MEMORY.md — Persistent long-term agent memory and user preference mapping
  • memory/YYYY-MM-DD.md — Daily event and interaction logs
  • memory/timeout-state.json — Session timeout and notification state

• Skills

  • skills/midscene-yaml/
    • SKILL.md — Midscene YAML automation skill, formats, and triggers
    • references/midscene-yaml-reference.md — Full syntax reference for skill
    • references/examples.md — Usage examples across platforms

• Docs & Training Materials

  • docs/复盘报告-2026-03-06至03-07.md — Postmortem review/report log
  • docs/德鲁克培训/ — Drucker management training package
    • generate_ppt.py — PPT automation script
    • 培训师手册.md, 培训提纲.md, 学员手册.md — Trainer/trainee materials

How to Use

1. Initialization

   • Read AGENTS.md and SOUL.md to understand agent role, etiquette, session flow, and core identity.
   • Load MEMORY.md and most recent memory/YYYY-MM-DD.md files for relevant user context and state.
   • Review or contribute to daily memory files and persistent logs as the main method of keeping agent context.

2. Agent Operation

   • Follow the etiquette and decision-making principles (see AGENTS.md).
   • Use persistent memory (write to MEMORY.md and daily memory files) for important events, lessons, and user preferences.
   • Refer to technical notes in MEMORY.md for Python compatibility fixes, CLI habits, and configuration guidance.

3. Skills Integration

   • Integrate and use the midscene-yaml skill for natural language to UI automation script conversion. See skills/midscene-yaml/SKILL.md and references for details.
   • For custom skills and extensions, add references and documentation into the appropriate /skills subfolders.

4. Multi-Agent & Tooling

   • If using multiple OpenClaw agent workspaces, follow directory and shared resource structures as outlined in TOOLS.md.
   • Utilize submodules for sharing skills among multiple agent workspaces if necessary.

5. Documentation & Training

   • Leverage the Drucker management training docs for onboarding, presentations, or creating custom agent-driven training material.

Notes

• Memory Model: Every session starts "fresh," but continuity is achieved via explicit file-based memory. Always document what must persist.
• Etiquette: Agent responses must observe user-preferred forms of address at all times; see MEMORY.md for mapping user IDs to titles/honorifics.
• Privacy & Safety: Sensitive configuration (such as tokens, credentials, or private context) should be managed securely and never disclosed.
• Technical Fixes: The workspace contains several workaround notes for Python 3.6 compatibility, Feishu integration, OpenClaw upgrades, and system hardening—refer to MEMORY.md and relevant daily memory logs.
• Expansion: For new agents or skills, follow the current documentation structure for consistency and maintainability.
• Language: Materials and logs are primarily in Chinese. Adapt or translate as appropriate for your environment.