← 返回 Skills 市场
565
总下载
0
收藏
1
当前安装
8
版本数
在 OpenClaw 中安装
/install bot
功能描述
The Universal Autonomous Entity Standard. A local-first framework for observable, composable agents with policy-guarded execution on ClawHub.
使用说明 (SKILL.md)
BOT: Standardized Agent Framework
bot is a local-first agent framework for observable execution, composable tool use, and multi-agent orchestration.
Safety Model
- Network: Core demos do not initiate outbound network requests by default.
- Execution: This release provides policy-guarded local execution, not OS-level sandbox isolation.
- File System: Local memory is stored in
./.bot_memory/by default, configurable viaBOT_MEMORY_DIR. - Viewer: Optional local monitor binds to
127.0.0.1only when explicitly started. - Credentials: No external API keys required for core demos.
Included Capabilities
- Single-agent reasoning core
- Botfile static declaration
- TUI thought visualization
- Crypto identity and signing
- Multi-agent coordination
- Local Web thought-tree viewer
- Unified local run entrypoint
- Capability-aware tool registration
- Policy-guarded execution layer
Quick Start
PYTHONPATH=. python3 examples/run_bot.py --mode multi --prompt "Design a safe local-first agent workflow"
Custom Roles
PYTHONPATH=. python3 examples/run_bot.py --mode multi --roles "planner_bot,critic_bot,executor_bot,auditor_bot"
Single-Agent Mode
PYTHONPATH=. python3 examples/run_bot.py --mode single --prompt "Summarize this task safely"
Then open:
- http://127.0.0.1:8765
安全使用建议
This skill is a local agent framework and appears coherent with its description, but it executes Python callables directly (PolicyExecutor.execute) and stores information under ./.bot_memory by default. Before running: (1) run inside a virtual environment (install.sh recommends this), (2) inspect any Botfile or third-party tools you register to ensure they don't perform unwanted filesystem or network actions, (3) be aware there is no OS-level sandboxing — policies are runtime checks only, not enforcement of OS restrictions, and (4) the web viewer binds to 127.0.0.1 by default (local only). If you plan to register or load untrusted code, treat it as potentially able to perform arbitrary local actions and isolate it accordingly.
功能分析
Type: OpenClaw Skill
Name: bot
Version: 2.1.0
The 'bot' skill bundle is a well-structured framework for local agent orchestration, featuring multi-agent coordination, policy-guarded execution, and a local visualization tool. Analysis of the source code (e.g., src/core/reasoning.py, src/protocol/alignment.py) and the local web server (src/web/thought_tree.py) shows no evidence of data exfiltration, malicious execution, or prompt injection; all operations are consistent with the stated goal of providing a safe, local-first agent environment.
能力评估
Purpose & Capability
Name/description (local-first agent framework) aligns with the delivered files: agent core, multi-agent coordinator, tool registry, policy layer, local web viewer, and examples. Declared requirements (none) are consistent with a pure-Python, local-first demo framework.
Instruction Scope
SKILL.md and example code are focused on local execution and explicitly state no outbound network by default. The runtime instructions and code read/write a local memory directory (./.bot_memory or BOT_MEMORY_DIR) and optionally load a Botfile. The code can execute arbitrary Python callables via PolicyExecutor.execute (no OS-level sandboxing) — this is expected for a framework but is a behavioral surface the user should understand and control (only register trusted tools).
Install Mechanism
No platform install spec is declared (instruction-only from registry), though an install.sh and requirements.txt are included for manual setup (pip install -r requirements.txt). No downloads from untrusted URLs or archive extraction are present.
Credentials
No required credentials or env vars are declared. The code optionally respects BOT_MEMORY_DIR to configure local memory storage. No secret-scoped environment variables are requested or used.
Persistence & Privilege
always is false and the skill does not request elevated platform privileges or modify other skills. It persists local agent memory under a local directory it creates (./.bot_memory) which is proportional to its purpose.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install bot - 安装完成后，直接呼叫该 Skill 的名称或使用
/bot触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v2.1.0
Version 2.1.0
- Introduced policy-guarded execution: agents now enforce policies during tool use and code execution.
- Added a policy executor module and a policy demo example to illustrate new safety controls.
- Improved documentation: clarified the safety model, updated quick start info, and detailed new capabilities in SKILL.md.
- Expanded agent features with capability-aware tool registration and support for crypto identity/signing.
- Removed legacy sandbox code in favor of the new policy-driven approach.
v2.0.1
Version 2.0.1 introduces a unified local run entrypoint for easier agent execution.
- Added examples/run_bot.py as the primary entrypoint for running agents in both single and multi-agent modes.
- Expanded documentation with unified quick start and usage examples for single and multi-agent workflows.
- Updated internal web viewer integration and usage instructions.
v2.0.0
Version 2.0.0 of bot introduces multi-agent orchestration and web-based visualization.
- Added support for multi-agent coordination and demo script (`examples/multi_agent_demo.py`).
- Introduced local web-based "thought tree" viewer (`src/web/thought_tree.py`).
- Expanded documentation to cover new features and quick start for multi-agent mode.
- Updated security model details and included observability/visualization improvements.
- Updated dependencies and instructions to reflect new capabilities.
v1.3.0
- Introduced cryptographic identity support by adding src/protocol/crypto_identity.py.
- Updated core reasoning logic in src/core/reasoning.py, possibly to utilize or integrate new identity features.
v1.2.0
Version 1.2.0
- Added a new TUI (Text User Interface) tool at src/tools/tui.py to enhance user interaction options.
- Updated core reasoning logic in src/core/reasoning.py.
v1.1.0
bot 1.1.0
- Added new file: src/tools/botfile.py
- Introduces initial support for bot file tools.
v1.0.1
BOT 1.0.1 — Foundational release for secure, composable autonomous agents.
- Added modular source files for perception, reasoning, alignment, identity, memory, tool registry, and sandboxing.
- Security: Network access is default-deny; file system sandboxed by default; runs locally without external API keys.
- Provided install script and requirements.txt for easy setup.
- New quick start Python example demonstrates basic agent creation and reasoning cycle.
- Updated documentation to reflect security features and usage on ClawHub.
v1.0.0
Initial sovereign entity release
元数据
常见问题
bot 是什么？
The Universal Autonomous Entity Standard. A local-first framework for observable, composable agents with policy-guarded execution on ClawHub. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 565 次。
如何安装 bot？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install bot」即可一键安装，无需额外配置。
bot 是免费的吗？
是的，bot 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
bot 支持哪些平台？
bot 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 bot？
由 agenticio（@agenticio）开发并维护，当前版本 v2.1.0。
推荐 Skills