← 返回 Skills 市场
144
总下载
0
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install agent-teacher
功能描述
教授新 agent 掌握工作室基础能力的技能。触发时机：(1) 创建新 agent 后需要初始化配置时 (2) 教现有 agent 掌握某项技能时 (3) 新 agent 上岗培训。课程分"行为准则→基础→进阶"三大类，行为准则必须优先于一切技能学习。
使用说明 (SKILL.md)
Agent Teacher
教授新 agent 掌握工作室基础能力的技能。
课程理念
- 准则优先：行为准则必须先于一切技能学习，准则比技能更重要
- 基础扎实：技能让 Agent 做事情，准则让 Agent 不犯错
- 循序渐进：按顺序完成课程，不要跳跃
- 不重复记录：已经学过并记住的内容不要重复记录，直接巩固现有知识点即可
课程体系
行为准则（必修第一课）
必须先于所有技能学习！ 准则违规比技能不熟练更严重。
| 课程 | 内容 |
|---|---|
rules-of-conduct |
工作室行为准则（约15条核心规则） |
详见 references/rules-of-conduct.md
身份定制（必修第1.5课）
在安装任何技能之前必须完成！ 定义你是谁、你的工作规范。这些文件必须中文编写。
| 课程 | 内容 |
|---|---|
phase-0-identity |
SOUL.md、IDENTITY.md、AGENTS.md 定制 |
详见 references/phase-0-identity.md
基础课程（必修）
在行为准则学习后进行。
| 分类 | 技能 | 安装 |
|---|---|---|
| 持续存在 | daily-log, memory-review, daily-backup, todo-list | clawhub 安装 |
| 沟通协作 | feishu-send, contacts, sessions_send | 部分安装部分内置 |
| 知识检索 | memory_search, memory_get | 内置 |
| 飞书文档 | feishu_doc/wiki/bitable | 内置 |
| 系统维护 | health-check, dependency-tracker | clawhub 安装 |
详见 references/phase-1-foundation.md
进阶课程（选修）
按需学习，用于专业任务。
| 分类 | 技能 | 安装 |
|---|---|---|
| 搜索 | mmx-cli, multi-search-engine | mmx-cli 优先，mcporter 备选 |
| 浏览器 | browser | 内置需启用 |
| 外部服务 | mcporter | 需配置key |
| 视觉 | canvas, image, mmx-cli | image 理解用内置工具，生图用 mmx-cli |
| 技能开发 | skill-creator, clawhub | 部分内置部分安装 |
详见 references/phase-2-advanced.md
新 agent 上岗教学流程
Step 1: 学习行为准则（必须第一个）
读取并理解 references/rules-of-conduct.md
→ 确认理解后再进行下一步
Step 2: 安装并学习基础技能
cd /home/axelhu/.openclaw/workspace/[agent-name]
mkdir -p skills
# 基础技能
clawhub install daily-log --dir skills
clawhub install memory-review --dir skills
clawhub install daily-backup --dir skills
clawhub install todo-list --dir skills
clawhub install feishu-send --dir skills
clawhub install health-check --dir skills
clawhub install dependency-tracker --dir skills
cp -r /home/axelhu/.openclaw/workspace/skills/contacts skills/
# sessions_send、memory_search、memory_get、feishu_doc/wiki/bitable、skill-creator 已内置
Step 3: 验证掌握
- 行为准则：能复述至少 5 条核心规则
- 基础技能：能发日报、能搜记忆、能给其他 agent 发消息
Step 4: 进阶按需学习
根据角色和任务需要，安装进阶技能。
详细课程资料
- 行为准则：
references/rules-of-conduct.md - 基础技能：
references/phase-1-foundation.md - 进阶技能：
references/phase-2-advanced.md
汇报模板
已完成 [agent-name] 上岗培训 ✅
行为准则：✅ 已掌握（能复述核心规则）
基础技能：✅ daily-log, memory-review, daily-backup, todo-list, feishu-send, contacts, health-check
进阶技能：按需安装
触发示例
- "教 zero-producer 上岗培训"
- "新 agent 上岗，按顺序教学"
- "给 programmer 补行为准则"
安全使用建议
Before installing or invoking this skill: 1) Confirm whether AGENTS.md is per-agent or a shared/global file — do NOT apply the SKILL.md's 'replace memory section' step to a global config without team approval. 2) Back up AGENTS.md, SOUL.md, IDENTITY.md and the workspace before making edits. 3) Ask the skill author (or inspect locally) for the exact effect of the recommended clawhub installs and where clawhub will fetch packages from. 4) Remove or adapt hardcoded paths (/home/axelhu) to your environment; test in a non-production/sandbox agent first. 5) Question the instruction that mandates making MEMORY.md accessible to all sessions: this broadens data exposure and may violate privacy rules — require explicit human authorization. 6) If you want to proceed, run the file-edit steps interactively (human-in-the-loop) rather than letting an agent autonomously apply the changes. 7) If you are unsure, mark this skill for manual onboarding only and have an engineer review the implications for memory access and contact-list copying.
功能分析
Type: OpenClaw Skill
Name: agent-teacher
Version: 1.3.1
The bundle is an administrative tool designed to automate the onboarding and configuration of AI agents. It possesses high-risk capabilities, including installing software via 'clawhub', modifying core agent configuration files (SOUL.md, AGENTS.md), and executing shell commands for workspace setup. While these actions align with the 'Agent Teacher' purpose, the instructions in SKILL.md exhibit potential shell injection vulnerabilities by using unsanitized placeholders like '[agent-name]' in commands. Furthermore, references/phase-1-foundation.md details the manual handling of Feishu API credentials via curl, which is a risky practice. No clear evidence of intentional malice or data exfiltration was found, but the broad permissions and lack of input validation in the instructions meet the criteria for a suspicious classification.
能力标签
能力评估
Purpose & Capability
The name/description (agent onboarding/teaching) matches the instructions to install foundation skills and teach rules. However some requirements in SKILL.md (editing AGENTS.md to change memory-access rules, forcing Chinese-only identity files, hardcoded /home/axelhu paths) go beyond simple 'teaching' and imply making policy/configuration changes that could have broader impact.
Instruction Scope
Runtime instructions tell the agent to read and modify workspace files (SOUL.md, IDENTITY.md, AGENTS.md) and to replace a 'Memory' section with a new global policy that makes MEMORY.md accessible across sessions. That is not just onboarding content: it is a configuration/policy change that affects how memory is shared and could expose data across sessions/agents. The SKILL.md also hardcodes absolute paths under /home/axelhu and runs clawhub install/cp commands that operate on the user's workspace.
Install Mechanism
This is an instruction-only skill (no install spec, no code files), which is low-risk by itself. The instructions do call system tooling (clawhub install, cp) to add other skills — expected for onboarding — but the skill doesn't declare where those packages come from or verify sources. That's typical but worth confirming (what clawhub will install and from which registry).
Credentials
The skill requests no env vars or credentials, which is good. However it requires read/write access to workspace files and explicit replacement of AGENTS.md memory rules — a privileged change relative to a purely educational skill. The hardcoded path /home/axelhu and copying from that location may be inappropriate for other environments and could lead to accidental cross-agent data copying (e.g., contacts).
Persistence & Privilege
always:false and normal autonomous invocation are fine. The main concern is that the instructions explicitly tell the agent to permanently modify AGENTS.md (delete restrictions and replace memory section) — a change that can persist system-wide and increase memory sharing across sessions/agents. The skill does not explicitly limit modifications to a single agent's isolated config, so this is a privilege escalation risk if AGENTS.md is shared.
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 agent-teacher - 安装完成后，直接呼叫该 Skill 的名称或使用
/agent-teacher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
v1.3.1
扩充mmx-cli详细用法（search/image/speech/video/music/vision全子命令）
v1.3.0
清理过时内容：删除项目关闭流程/更新canvas规范/更新memory路径示例
v1.0.1
删除过时内容：项目关闭流程/canvas action规范/memory路径示例
v1.2.0
新增 minimax web_search 教程，补充中文搜索优先指引
元数据
常见问题
Agent Teacher 是什么？
教授新 agent 掌握工作室基础能力的技能。触发时机：(1) 创建新 agent 后需要初始化配置时 (2) 教现有 agent 掌握某项技能时 (3) 新 agent 上岗培训。课程分"行为准则→基础→进阶"三大类，行为准则必须优先于一切技能学习。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 144 次。
如何安装 Agent Teacher？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-teacher」即可一键安装，无需额外配置。
Agent Teacher 是免费的吗？
是的，Agent Teacher 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Agent Teacher 支持哪些平台？
Agent Teacher 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 Agent Teacher？
由 AxelHu（@axelhu）开发并维护，当前版本 v1.3.1。
推荐 Skills