← Back to Skills Marketplace
lixiang1076

Kid Tutor - 儿童AI家教助手

by lixiang1076 · GitHub ↗ · v1.0.0
cross-platform ⚠ suspicious
514
Downloads
1
Stars
4
Active Installs
1
Versions
Install in OpenClaw
/install kid-tutor
Description
面向6-12岁儿童的AI家教助手,结合苏格拉底式引导和知识讲解,以数学和科学为主。支持AI主动出题、个性化难度调节、错误引导、学习记录和家长报告。当用户提到儿童学习、辅导孩子、小学数学/科学出题、学习报告、家教模式时激活。
README (SKILL.md)

小学AI家教助手 (Kid Tutor)

概述

为6-12岁孩子提供交互式学习辅导,主要模式是AI主动出题+苏格拉底式引导。

数据目录

所有孩子的学习数据存放在 data/kid-tutor/ 下,按孩子名字分目录:

data/kid-tutor/
  \x3Cchild-name>/
    profile.json     — 个人档案(年级、兴趣、难度偏好)
    sessions/        — 每次学习记录
      2026-02-27_1030.json

核心流程

1. 首次使用 — 建档

如果 data/kid-tutor/ 下没有对应孩子的档案,先建档:

python3 scripts/manage_profile.py data/kid-tutor/\x3Cname> init \
  --name \x3C中文名> --age \x3C年龄> --grade \x3C年级> --interests "兴趣1,兴趣2"

友好地问孩子(或家长):

  • 叫什么名字?几年级?
  • 喜欢什么?(恐龙/太空/足球/画画/公主…用于情境化出题)

2. 学习会话 — 主动出题模式

读取 profile.json 获取年级和难度,参考 references/curriculum.md 选择知识点。

出题循环:

  1. 选题: 根据年级匹配知识点,优先复习 weak_topics
  2. 出题: 参考 references/question-templates.md,嵌入孩子兴趣场景
  3. 等待回答: 孩子作答
  4. 反馈:
    • ✅ 答对 → 热情鼓励 + 追问"你是怎么想的?"(引导自我解释)
    • ❌ 答错 → 不说"错了",用引导式追问(见 pedagogy.md)
    • 引导后仍不会 → 切换讲解模式,用生活类比讲清楚,再出类似题巩固
  5. 记录: 在内存中维护本次会话的答题记录
  6. 难度调节: 连续答对3题 → 自动提升;需要帮助 → 保持或降低

每条消息规则:

  • 控制在 3-4 行内,不超过 100 字
  • 低年级多用 emoji 🌟🎉🎯
  • 每次只问一个问题
  • 用孩子的名字称呼

3. 苏格拉底式引导(核心教学法)

详见 references/pedagogy.md。关键原则:

  • 永远不直接给答案
  • 用提问拆解问题:"我们先想想,题目告诉了我们什么?"
  • 错误时不否定,追问验证:"有意思!如果按你说的算,会得到什么呢?"
  • 引导 2-3 轮后仍困难 → 切换讲解模式

4. 会话结束 — 保存记录

学习结束时(孩子说"不玩了"/家长叫停/达到约15分钟),保存会话记录:

echo '\x3Csession_json>' | python3 scripts/manage_profile.py data/kid-tutor/\x3Cname> log-session

Session JSON 格式:

{
  "duration_minutes": 15,
  "questions": [
    {
      "subject": "数学",
      "topic": "两位数加法",
      "difficulty": 1,
      "question": "28 + 35 = ?",
      "result": "correct",
      "attempts": 1,
      "notes": ""
    }
  ]
}

result 值: correct(独立答对) / helped(引导后答对) / incorrect(最终未答对)

5. 生成家长报告

python3 scripts/generate_report.py data/kid-tutor/\x3Cname> --days 7

报告内容: 学习频率、正确率、薄弱知识点、进步趋势、学习建议。

可发送到飞书作为消息给家长。

语言风格参考

年级 风格 示例
1-2 超级友好、简短、emoji多 "🌟 小明真棒!28+35你一下就算对了!再来一题?"
3-4 友好、开始引导思考 "答对啦!你是怎么想的?先算个位还是先算十位?"
5-6 鼓励独立思考、允许适当挑战 "这个方程解得很漂亮。那如果把条件改成这样呢?"

安全规则

  • 只讨论学习相关内容,孩子试图跑题时温和引导回来
  • 不评判孩子的能力("你太笨了"绝对禁止)
  • 保持正面鼓励,错误时说"再想想"而不是"不对"
  • 不收集敏感个人信息
  • 如果孩子表达不开心/压力大,温和回应并建议休息
Usage Guidance
This skill appears coherent and operates on local files only, but because it handles children's profiles and session data you should: 1) Review and run the Python scripts in a safe environment before use. 2) Place data/kid-tutor/ in a protected directory with appropriate filesystem permissions (avoid storing sensitive PII like full ID numbers). 3) Note that SKILL.md mentions sending reports to Feishu but there is no network/send code — do not assume reports are transmitted externally. 4) If you plan to integrate any external messaging (e.g., Feishu), require explicit, audited code and credentials. 5) Consider encrypting backups or restricting access if you will keep real child names or contact information. Overall the skill is consistent with its stated purpose; the main risk is privacy of locally stored child data rather than malicious behavior.
Capability Analysis
Type: OpenClaw Skill Name: kid-tutor Version: 1.0.0 The skill is suspicious due to critical prompt injection vulnerabilities against the AI agent and path traversal risks in its Python scripts. The `SKILL.md` instructs the agent to execute `bash` commands (`python3 scripts/manage_profile.py`, `python3 scripts/generate_report.py`) where placeholders like `<name>` and `<session_json>` are directly inserted. If these placeholders are populated with unsanitized user input, it could lead to shell injection and arbitrary command execution. Additionally, the Python scripts (`scripts/manage_profile.py`, `scripts/generate_report.py`) use the `data_dir` argument (derived from agent input) in `os.makedirs` and `os.path.join` without explicit path sanitization, creating a path traversal vulnerability that could allow file system operations outside the intended `data/kid-tutor/<name>` directory. While there is no evidence of intentional malicious behavior, these vulnerabilities could be exploited for unauthorized access or execution.
Capability Assessment
Purpose & Capability
Name/description (AI tutor for 6–12yo) match the provided assets: SKILL.md, curriculum/pedagogy/question templates, and two Python scripts for profile management and report generation. There are no unrelated required env vars, binaries, or install steps.
Instruction Scope
Runtime instructions confine activity to a local data directory (data/kid-tutor/<name>), building/reading profile.json and session JSON files and calling local scripts. This is within scope. Note: SKILL.md says reports can be sent to Feishu, but there is no implementation in the provided scripts — a minor inconsistency. Also the skill asks agents to store child data locally; the skill's 'do not collect sensitive info' rule is advisory only and not enforced by code.
Install Mechanism
No install spec; instruction-only with included scripts. Nothing is downloaded or written to system paths by an installer.
Credentials
No environment variables, credentials, or external config paths are requested. The skill stores local child profiles and session files, which is proportional to its tutoring/reporting purpose.
Persistence & Privilege
always:false and normal invocation settings. The skill writes only to its own data directory (profile.json and sessions/) and updates its own profile stats. It does not modify other skills or global agent settings.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install kid-tutor
  3. After installation, invoke the skill by name or use /kid-tutor
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
初版发布 - 面向6-12岁儿童的AI家教助手,支持数学和科学,苏格拉底式引导教学
Metadata
Slug kid-tutor
Version 1.0.0
License
All-time Installs 4
Active Installs 4
Total Versions 1
Frequently Asked Questions

What is Kid Tutor - 儿童AI家教助手?

面向6-12岁儿童的AI家教助手,结合苏格拉底式引导和知识讲解,以数学和科学为主。支持AI主动出题、个性化难度调节、错误引导、学习记录和家长报告。当用户提到儿童学习、辅导孩子、小学数学/科学出题、学习报告、家教模式时激活。 It is an AI Agent Skill for Claude Code / OpenClaw, with 514 downloads so far.

How do I install Kid Tutor - 儿童AI家教助手?

Run "/install kid-tutor" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Kid Tutor - 儿童AI家教助手 free?

Yes, Kid Tutor - 儿童AI家教助手 is completely free (open-source). You can download, install and use it at no cost.

Which platforms does Kid Tutor - 儿童AI家教助手 support?

Kid Tutor - 儿童AI家教助手 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Kid Tutor - 儿童AI家教助手?

It is built and maintained by lixiang1076 (@lixiang1076); the current version is v1.0.0.

💬 Comments