← 返回 Skills 市场
171
总下载
0
收藏
1
当前安装
16
版本数
在 OpenClaw 中安装
/install kaiwu8
功能描述
开悟吧平台型 Skill - 管理并激活已购功能。用户说"开悟吧"时执行 activate.py 脚本。
使用说明 (SKILL.md)
开悟吧 kaiwu8
用户只需要安装这一个 skill，所有功能从这里激活
触发条件
用户说："开悟吧"
执行步骤
- 运行 activate.py
- 系统自动调 API 检查已购功能
- 未激活的功能自动执行激活
- 缺失的 skill 从 ClawHub 下载
配置
在 openclaw.json 的 skills.entries.kaiwu8 中配置：
"kaiwu8": {
"api_endpoint": "https://api.kaiwu8.com",
"user_api_key": "你的user_api_key"
}
安全使用建议
This skill will contact a remote API (api_endpoint) using a user_api_key you must provide (in openclaw.json or via environment variables), may install other skills by invoking the 'clawhub' CLI, and will modify all agents.md files under ~/.openclaw/workspace/agents-soul based on content returned from the server. Before installing: (1) confirm the api_endpoint and the trustworthiness of the provider; (2) back up your workspace/agents.md files; (3) inspect the server API responses (what/why/how) or run the script in a safe/sandboxed environment first; (4) ensure you are comfortable with the script having permission to run 'clawhub' and to change multiple agents' configurations; (5) ask the publisher to update the skill metadata to declare the required config/env vars and to document exactly what is sent to the server and what payloads might be written into agents.md. If you cannot validate the remote endpoint and payloads, treat this skill as high-risk and avoid enabling it on production agents.
功能分析
Type: OpenClaw Skill
Name: kaiwu8
Version: 1.5.0
The skill 'kaiwu8' implements a remote activation mechanism that fetches instructions from an external API (default: api.kaiwu8.com) and programmatically modifies the system prompts (agents.md) of all local agents via activate.py. This allows a remote server to perform prompt injection or alter agent logic across the entire workspace. Additionally, the script uses subprocess.run to install further skills from ClawHub based on remote API responses, creating a significant supply-chain risk and remote control vector, although these actions are aligned with the stated purpose of managing 'purchased features'.
能力标签
能力评估
Purpose & Capability
Name/description claim to manage and activate purchased features; the code implements that (reads config, queries a remote API for purchased features, installs missing skills from ClawHub, activates features). However the registry metadata declares no required env/config, while both SKILL.md and the code require an api_endpoint and a user_api_key to be present in openclaw.json (or via KAIWU8_API_ENDPOINT / KAIWU8_USER_API_KEY). This mismatch is a coherence issue.
Instruction Scope
The runtime will (a) call the remote API with your user_api_key, (b) install missing skills by running the 'clawhub install' CLI, and (c) find and modify all agents.md files under ~/.openclaw/workspace/agents-soul according to server-provided what/why/how payloads. Allowing remote payloads to be directly appended/inserted into agents.md is high-impact and can change agent behavior arbitrarily; the SKILL.md wording ('Xiaobai 理解 what/why/how 并精准修改 agents.md') is open-ended and grants wide discretion.
Install Mechanism
There is no install spec (instruction-only + bundled activate.py). The script invokes an external CLI ('clawhub') via subprocess to download/install skills. This is expected for a skill installer but does require the 'clawhub' binary to exist and gives the script the ability to run shell commands (and cause additional code to be installed).
Credentials
Registry data lists no required environment variables or primary credential, but the script expects configuration either in ~/.openclaw/openclaw.json under skills.entries.kaiwu8 or via KAIWU8_API_ENDPOINT / KAIWU8_USER_API_KEY environment variables. Asking for a user_api_key (sensitive credential) is proportionate to contacting the remote activation API, but the fact these are not declared in the skill metadata is an inconsistency. Also the script reads openclaw.json from disk (which may contain other workspace config), increasing the surface area if the code were modified.
Persistence & Privilege
always is false (good), but the script writes state in ~/.openclaw/kaiwu8-state and logs in the skill directory. More importantly, it will modify many agents.md files across the user's workspace and can install other skills system-wide via clawhub. Those abilities provide a large blast radius (modify agent behavior, pull additional code) and should be treated as high privilege even though the skill is not force-enabled.
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 kaiwu8 - 安装完成后，直接呼叫该 Skill 的名称或使用
/kaiwu8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
v1.5.0
新增：环境变量 fallback（KAIWU8_API_ENDPOINT / KAIWU8_USER_API_KEY）
v1.4.0
修复：UTF-8 BOM 解码错误 + emoji GBK 崩溃
v1.3.0
新增：安装skill时带指数退避重试，避免被ClawHub限速卡死
v1.2.0
修复：先查后补逻辑，激活过的功能也检查skill安装状态
v1.1.1
优化提示语：找不到 agents.md 时直接告诉用户把内容加到 agents.md，不用说「复制」
v1.1.0
新增：找不到 agents.md 时，打印内容让用户手动复制（而不是直接失败），确保所有边界情况都能完成激活
v1.0.9
修复：find_all_agents_md() 写死了彧哥的目录结构，改为通用扫描（优先 agents-soul 目录，找不到则递归扫 ~/.openclaw/ 下所有 agents.md）；找不到时给出明确路径提示
v1.0.8
修复：(1) 换 requests 库解决 Python urllib 401 问题；(2) 新增 rule_content 模式，服务端返回 rule_content 时直接追加到 agents.md
v1.0.7
修复：Windows GBK 终端 emoji 打印崩溃，所有 print 改为 _safe_print 兼容处理
v1.0.6
新增：activate.py 启动时自动检查并更新到最新版本，用户无需手动跑 update 命令
v1.0.5
文档更新：配置改用环境变量，移除 openclaw.json 说明（当前版本有 schema bug）
v1.0.4
修复：支持环境变量 KAIWU8_API_ENDPOINT + KAIWU8_USER_API_KEY，绕过 openclaw.json schema 校验
v1.0.3
新架构：what/why/how + ClawHub下载skill + Xiaobai理解执行，零手动激活
v1.0.2
test
v1.0.1
kaiwu8 v1.0.1
- 精简和更新了 SKILL.md 说明，只保留基本描述与触发方式
- 移除了 _meta.json、activated.json 和 config.json 等配置及本地状态文件
- 修改 activate.py 实现以配合去除后的文件与精简流程
v1.0.0
Initial release: 开悟吧 skill 订阅平台用户端
元数据
常见问题
开悟吧 是什么？
开悟吧平台型 Skill - 管理并激活已购功能。用户说"开悟吧"时执行 activate.py 脚本。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 171 次。
如何安装 开悟吧？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install kaiwu8」即可一键安装，无需额外配置。
开悟吧 是免费的吗？
是的，开悟吧 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
开悟吧 支持哪些平台？
开悟吧 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 开悟吧？
由 MarkMa84（@markma84）开发并维护，当前版本 v1.5.0。
推荐 Skills