← Back to Skills Marketplace
314
Downloads
1
Stars
0
Active Installs
4
Versions
Install in OpenClaw
/install multi-windows
Description
多窗口工作台，管理多个独立工作窗口，切换时可自动保存和恢复工作进度
Usage Guidance
This skill operates on local OpenClaw workspace and session files (paths under ~/.openclaw). Before installing, be aware it will read your agent session transcripts (latest .jsonl in ~/.openclaw/agents/main/sessions) and write metadata and transcript/summary files under ~/.openclaw/workspace/memory/tasks. That behavior is necessary for saving/restoring windows but means the skill can access your conversation history stored locally. If you are comfortable with that, the code is coherent and there are no network calls or external credential requirements. If you have sensitive transcripts you do not want consolidated, do not install or relocate/secure your session directory; alternatively review the scripts locally and modify paths/permissions as needed.
Capability Analysis
Type: OpenClaw Skill
Name: multi-windows
Version: 1.0.3
The skill bundle provides a local workspace management system for the OpenClaw agent, allowing it to organize tasks into 'windows' with isolated contexts. The Python scripts (e.g., switch.py, create.py) perform standard file I/O operations within the ~/.openclaw directory to manage task metadata and session transcripts. No evidence of data exfiltration, network activity, or malicious intent was found; the access to session history is consistent with the stated goal of maintaining conversation context across different tasks.
Capability Assessment
Purpose & Capability
Name/description (多窗口模式 — manage windows and save/restore progress) align with required files and behavior: scripts create/list/switch/archive/complete windows and read/write window metadata and transcripts in ~/.openclaw/workspace/memory/tasks and session files under ~/.openclaw/agents/main/sessions.
Instruction Scope
SKILL.md and scripts instruct the agent to read and write local workspace and session files (index tasks.json, current.json, meta.json, transcript.jsonl). This is consistent with 'save/restore conversation' functionality. Note: the skill reads the agent's session directory (~/.openclaw/agents/main/sessions) to find the latest .jsonl session file — this means it can access other locally stored session transcripts, which is expected for the feature but worth knowing.
Install Mechanism
No install spec; this is an instruction+script bundle (no downloads, no external installers). The code is included as scripts, so nothing is fetched from remote during install.
Credentials
No environment variables, credentials, or external config paths are requested. The scripts only access repository-defined local paths under the user's home (~/.openclaw/...). The requested access is proportional to the stated functionality.
Persistence & Privilege
Flags: always=false and agent autonomous invocation defaults unchanged. The skill writes and updates files within its own workspace directories (tasks, current, archived). It does not modify other skills or system-wide agent configuration outside ~/.openclaw/workspace/memory/tasks and writes a current.json there — behaviour is within expected scope.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install multi-windows - After installation, invoke the skill by name or use
/multi-windows - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.3
- Improved window context isolation: each window now saves and restores its own transcript.jsonl during switching.
- When switching windows, current session history is auto-saved; switching to a window displays its previous chat history.
- Refined documentation with explicit changelog, new usage examples, and updated directory structure.
- Core features and command references streamlined for clarity.
v1.0.2
- Added "归档窗口 / Archive Window" feature for archiving completed windows.
- New script archive.py introduced for the archive functionality.
- Updated documentation to describe archive commands and how to list archived windows.
- "归档窗口" commands now hide archived windows from the main list, improving window management.
v1.0.1
- 新增“保存为窗口”功能：可将当前对话一键独立保存为新窗口（新增脚本 save_as_window.py）。
- 新增“保存摘要”命令的独立实现（新增脚本 save_summary.py）。
- 文档全面优化：重新梳理命令示例、使用说明和文件结构，增加更清晰的可用指引和目录说明。
- 英中双语命令描述简化，提升易用性与可读性。
- 适配多窗口管理应用场景，细化窗口概念与操作流程说明。
v1.0.0
Initial release of multi-window workspace skill.
- Manage multiple independent work windows with context isolation and session switching.
- Supports creating, switching, listing, querying, and completing windows.
- Automatic context save and restore when switching windows.
- Each window has its own session and easy-to-remember ID.
- Commands available in both Chinese and English.
- Shows recent work summary and allows updating summaries per window.
Metadata
Frequently Asked Questions
What is 多窗口模式?
多窗口工作台，管理多个独立工作窗口，切换时可自动保存和恢复工作进度. It is an AI Agent Skill for Claude Code / OpenClaw, with 314 downloads so far.
How do I install 多窗口模式?
Run "/install multi-windows" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is 多窗口模式 free?
Yes, 多窗口模式 is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does 多窗口模式 support?
多窗口模式 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created 多窗口模式?
It is built and maintained by philonis (@philonis); the current version is v1.0.3.
More Skills