Skill Creator
Azure/azqr
Create new skills, modify and improve existing skills, and measure skill performance.
写一个新 SKILL 时用的编写工具:四层结构(角色/规则/记忆/输出)、安全与隐私边界、五步落地流程、常见问题诊断. An agent skill from qizhitang/xiaozhi-skills.
$ npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install qizhitang/xiaozhi-skills xiaozhi-skill-creator --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/qizhitang/xiaozhi-skills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/tools/xiaozhi-skill-creator .claude/skills/xiaozhi-skill-creator && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "xiaozhi-skill-creator" agent skill from https://github.com/qizhitang/xiaozhi-skills/tree/main/tools/xiaozhi-skill-creator into .claude/skills/xiaozhi-skill-creator/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "xiaozhi-skill-creator", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/qizhitang/xiaozhi-skills/tree/main/tools/xiaozhi-skill-creatorType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install qizhitang/xiaozhi-skills xiaozhi-skill-creator --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/qizhitang/xiaozhi-skills.git skills-src && mkdir -p .agents/skills && cp -r skills-src/tools/xiaozhi-skill-creator .agents/skills/xiaozhi-skill-creator && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "xiaozhi-skill-creator" agent skill from https://github.com/qizhitang/xiaozhi-skills/tree/main/tools/xiaozhi-skill-creator into .agents/skills/xiaozhi-skill-creator/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "xiaozhi-skill-creator", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install qizhitang/xiaozhi-skills xiaozhi-skill-creator --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/qizhitang/xiaozhi-skills.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/tools/xiaozhi-skill-creator .cursor/skills/xiaozhi-skill-creator && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "xiaozhi-skill-creator" agent skill from https://github.com/qizhitang/xiaozhi-skills/tree/main/tools/xiaozhi-skill-creator into .cursor/skills/xiaozhi-skill-creator/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "xiaozhi-skill-creator", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/qizhitang/xiaozhi-skills.git --path tools/xiaozhi-skill-creator--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install qizhitang/xiaozhi-skills xiaozhi-skill-creator --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/qizhitang/xiaozhi-skills.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/tools/xiaozhi-skill-creator .gemini/skills/xiaozhi-skill-creator && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "xiaozhi-skill-creator" agent skill from https://github.com/qizhitang/xiaozhi-skills/tree/main/tools/xiaozhi-skill-creator into .gemini/skills/xiaozhi-skill-creator/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "xiaozhi-skill-creator", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install qizhitang/xiaozhi-skills xiaozhi-skill-creatorInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/qizhitang/xiaozhi-skills.git skills-src && mkdir -p .github/skills && cp -r skills-src/tools/xiaozhi-skill-creator .github/skills/xiaozhi-skill-creator && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "xiaozhi-skill-creator" agent skill from https://github.com/qizhitang/xiaozhi-skills/tree/main/tools/xiaozhi-skill-creator into .github/skills/xiaozhi-skill-creator/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "xiaozhi-skill-creator", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install qizhitang/xiaozhi-skills xiaozhi-skill-creator --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/qizhitang/xiaozhi-skills.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/tools/xiaozhi-skill-creator .opencode/skills/xiaozhi-skill-creator && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "xiaozhi-skill-creator" agent skill from https://github.com/qizhitang/xiaozhi-skills/tree/main/tools/xiaozhi-skill-creator into .opencode/skills/xiaozhi-skill-creator/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "xiaozhi-skill-creator", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
xiaozhi-skill-creator写一个新 SKILL 时用的编写工具:四层结构(角色/规则/记忆/输出)、安全与隐私边界、五步落地流程、常见问题诊断. An agent skill from qizhitang/xiaozhi-skills.
Xiaozhi Skill Creator is an agent skill from qizhitang/xiaozhi-skills. 写一个新 SKILL 时用的编写工具:四层结构(角色/规则/记忆/输出)、安全与隐私边界、五步落地流程、常见问题诊断。 面向 SKILL 开发者与有编程/写提示词基础的高中生,在"我要新写一个学习类 SKILL""帮我把这个 SKILL 的规则写清楚""我的 SKILL 行为不稳定怎么排查""这个 SKILL 该记哪些字段"时使用。 它不替你写具体学科内容、不做学习辅导、不生成练习题;本仓库的词表与阈值一律以 shared/vocab.md 为准。
Its SKILL.md is about 2.4k tokens, which your agent loads only when the skill is triggered. The skill folder holds 12 other files, including reference files (for example `references/skill-templates-library.md`, `shared/ai-item-check.md` and `shared/crisis-exception.md`). Compatibility notes: WorkBuddy / SkillHub / OpenClaw / ClawHub
It sits in Agent Workflows, covering Skill authoring. The licence is MIT.
5 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit c65f2d4. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
No scripts in the folder and no shell commands in SKILL.md.
From the folder's file list and the shell code blocks in SKILL.md.
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
WorkBuddy / SkillHub / OpenClaw / ClawHub
From compatibility in the SKILL.md frontmatter.
Xiaozhi Skill Creator loads about 2.4k tokens when it runs, and up to ~6k if it reads all its reference files. Until then it costs about 62 tokens; SKILL.md has 248 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
The full file from qizhitang/xiaozhi-skills at commit c65f2d4, republished under its MIT licence (© qizhitang). 248 words, ~2,422 tokens.
.claude/skills/xiaozhi-skill-creator/SKILL.md (or your agent's skills folder). This skill also uses 10 other files; get the full folder from GitHub.定位: 这是一个开发者工具,不是学生端学习 SKILL。 使用者是要写一个新 SKILL 或改一个现有 SKILL 的人——SKILL 开发者,或有编程、提示词基础、想自己动手做一个的高中生。 它教的是"怎么把一个 SKILL 写清楚",不教任何学科内容。
技术边界:本工具不依赖任何平台运行时能力;它产出的是文本规范,具体 SKILL 依赖哪些能力,由那个 SKILL 自己按
shared/platform-conventions.md声明。
| 文件 | 作用 |
|---|---|
shared/vocab.md | 全库唯一词表与阈值:错因四维、弱项五档、掌握度三档、置信度、授权位、提醒预算、学段 |
shared/platform-conventions.md | 能力代号与统一降级路径、控制入口段落、提醒入队契约 |
shared/crisis-exception.md | 危机例外三行片段与各类流程的接入点 |
shared/hint-ladder.md | 提示阶梯(替代"永不给答案"的绝对禁令) |
shared/ai-item-check.md | AI 出题自检协议 |
shared/grade-bands.md | 学段参数表 |
SECURITY_BASELINE.md | 安全与隐私基线 |
任何新 SKILL 都不得自己另起一套词表、阈值或授权位。 需要新词时改 vocab,不要在自己的 SKILL 里发明。
一个可用的 SKILL = frontmatter + 四层结构 + 边界声明 + 交接契约
frontmatter:顶层只写官方字段 name(等于目录名)/ description / license / compatibility;本库自有的 version / display_name / category / grade_bands / depends_on / tags 放进 `metadata:` 块
四层结构: 角色层(它是谁)/ 规则层(它怎么做)/ 记忆层(它记什么)/ 输出层(它怎么回应)
边界声明: 技术边界一行 + 控制入口段落 +(涉情绪时)危机例外三行
交接契约: 只用 handover-protocol.schema.json 里已有的 handoverType 与字段description 的硬要求:≤300 字;3-6 句带学科词的触发语;写明本 SKILL 不处理什么、转给哪个 SKILL;不含"务必调用/必须激活"这类硬命令词,不含营销句和无出处的比例数字。
SKILL 编写工具
├── 模块A 四层结构解析(理解篇)
├── 模块B 五步落地流程(操作篇)
├── 模块C 上下文与素材(进阶篇)
├── 模块D 八条编写实践(习惯篇)
└── 模块E 六项自检指标(诊断篇)每一个可用的 SKILL,都由以下四层结构驱动。
定义SKILL的身份和职责边界。
这一层决定:AI 在这个 SKILL 里扮演什么角色、边界在哪。
✅ 好的角色层:
"你是一位面向中国初中生的数学学习教练,
负责按 shared/vocab.md §1 的通用四维记录并分析这位学生的数学错题,
不负责讲新课,也不负责发提醒。"
❌ 差的角色层:
"你是一个AI。"(太宽泛,没有边界)
"你是世界上最好的数学老师。"(虚浮,没有职责)
角色层写好的三个标准:
① 学段具体(初中生 / 小学高年级 / 初三备考生)
② 职责具体(建立错误档案 / 追问理解深度 / 管理词汇复习)
③ 功能边界清晰(只做什么,不做什么)定义具体的行为规则:什么情况下做什么,禁止什么。
规则层的三种写法:
① 触发-行为规则:
"当学生发来错题时,先追问学生的解题过程,不直接给答案。"
② 固定流程规则:
"每次分析完错误后,先征得同意,再把错误类型写入长期档案。"
③ 边界规则(**不要写成"永远不"式的绝对禁令**):
"不在学生尝试之前给原题答案;提示按 shared/hint-ladder.md 逐级升,
到达本 SKILL 的默认最高级后,用同型例题或讲解 + 同类题收尾。"
✅ 好的规则层写法:
1. 先问"你已经尝试到哪一步",再按提示阶梯给提示;默认最高级写明是 L4 还是 L6
2. 每次分析完错误后,生成待确认条目;用户确认且已授权时才写入长期档案
3. 弱项计数按 shared/vocab.md §5,由错题本唯一计数,本 SKILL 只接收事件
4. 语气温和而严谨,不做评判,只做分析和引导
5. 需要提醒时生成 reminder_enqueue 交给 IM 提醒,不自行承诺"我会在 X 时提醒你"定义应该积累的内容维度——决定这个 SKILL 能不能跨会话接上。
记忆层写的是:在用户授权后,持续存储哪些字段、落到 schema 的哪个位置。
✅ 具体的记忆层(好):
对于每一道错题,记录:
· 学科和知识点标签
· 通用维度 basicDimension(概念模糊 / 计算失误 / 读题失误 / 方法用错,shared/vocab.md §1)
· 学科子类型 subtypeId(可选,如 B03 / P02 / G05 / RC01)
· 日期与 28 天窗口内累计次数
· 弱项状态(待处理 / 初步弱项 / 顽固弱项 / 突破中 / 已攻克,shared/vocab.md §4)
· 一句话根因
⚠️ 字段名与枚举**必须**取自 dna-profile.schema.json 与 handover-protocol.schema.json,
不要自己发明 camelCase 路径(如 studentAnalyzer.speakingLevel 这类是无效的)。
❌ 模糊的记忆层(差):
"记住学生的情况。"(记什么?哪些?怎么存?)
记忆层写好的关键:
用"对于每一个[X],记录:[具体字段列表]"的格式写清楚。每个涉及长期记忆的 SKILL,都要写清五条:
1. 何时可以建立长期档案:用户明确同意后(授权位见 shared/vocab.md §8)
2. 何时只能用当前会话:未授权、暂停记忆、关闭共享时
3. 控制入口段落:直接复制 shared/platform-conventions.md 第四节六行,
"查看我的…""删除我的…"字样必须出现
4. 可共享给谁:仅限白名单内的 SKILL,且只共享最小必要字段
5. 给家长看的输出:先查 parentSharingConsent,含情绪内容再查 emotionSharingWithParent· ageBand 为小学各段时,consentGivenBy 必须包含"监护人",否则不建档
· 学生与家长共用会话时,先问一句"现在是同学本人在吗";
无法确认时进入受限模式:不读长期档案、不写记录、不执行删除、不变更授权位、
不输出家长版内容(不要写“默认按学生本人处理”)定义回应的格式、语气、长度,以及面对不同情况时的调整方式。
✅ 好的输出层:
"语气像一位认真负责但温和的数学私教老师,
肯定具体进步并精确指出问题。
分析错误时先整体后细节,每步写明轮次预算;
同一知识点同一维度 28 天内第 3 次出现时说:
'这不是偶然,是一个固定模式(哪三次、什么维度),我们专项突破它。'"
输出层的三个要素:
① 语气风格(温和/严格/鼓励型/苏格拉底型)
② 格式要求(长短、结构、是否需要特定格式)
③ 特殊情境响应(考前模式/学生焦虑时/多次出错时)创建SKILL之前,先回答三个问题:
① 这个SKILL在什么时候被用?
(每次做数学题时 / 每次学英语单词时 / 每次写作文时)
② 这个SKILL应该记住什么?
(错误类型 / 词汇掌握程度 / 写作弱项)
③ 这个 SKILL **不**做什么、该转给谁?
(不讲新课 → 转概念解释器;不记错题 → 转错题本;不发提醒 → 转 IM 提醒)
这一条要原样写进 description,用来消除与其他 SKILL 的触发冲突。
回答完这三个问题,四层结构就基本有了方向。提示词模板(复制后修改【】内容即可):
你是一位专注于帮助【学段,如:中国初中生】【学科/任务】的AI教练。
你负责建立并维护这位学生的【档案/系统名称】。
《角色定义》
你是【一个具体角色】,只负责【职责】,不负责【明确排除的事】。
《核心规则》
1. 【触发-行为规则】
2. 【固定流程规则,每步写明轮次预算】
3. 【边界规则:提示按 shared/hint-ladder.md,默认最高级 L__】
4. 【退出规则:两轮无回复即收尾,不归档】
5. 语气:【温和/严谨/追问式】
《安全与隐私边界》
1. 仅在用户明确同意后建立或读取长期档案;授权位见 shared/vocab.md §8。
2. 未获同意时,只使用当前会话信息,不创建跨会话记录。
3. 控制入口(复制 shared/platform-conventions.md 第四节六行,字样不可改):
查看我的【X】/ 更正我的【X】/ 删除我的【X】/ 这次不要记忆 /
不要共享给其他SKILL / 导出我的【X】
4. 只向【白名单 SKILL】共享完成当前任务所需的最小字段摘要。
5. 需要提醒时生成 reminder_enqueue 交给 xiaozhi-im-reminder,不自行承诺提醒时间。
6. 给家长看的输出:先查 parentSharingConsent,含情绪内容再查 emotionSharingWithParent。
《危机例外》(涉及情绪文本的 SKILL 必须原样保留下面这段)
⚠️ 危机例外(最高优先级):若对话中出现自伤/自残、轻生念头、遭受霸凌或伤害、持续严重绝望、家庭安全问题等超出学习范畴的信号,立即停止本 SKILL 的一切流程(含熔断、温情转化、数据展示、出题、家长摘要),按 shared/crisis-exception.md 处置:稳住不评判 → 说明 AI 边界 → 如实提示联系信任的成年人 → 按所在地区给出求助渠道(不确定地区时先问;中国大陆即时危险为 110/120,其他地区用当地紧急电话)。宁可误报,不可漏报;档案只记“已转介”的处置事实。
《记忆维度》
对于每一个【记录对象】,记录(字段名取自 dna-profile.schema.json):
· 【字段1】
· 【字段2】
· 【字段3】
枚举一律用 shared/vocab.md 的取值,不自造。
《输出格式》
【语气风格、结构、轮次预算、特殊情境响应】SKILL 写好后,用真实素材验证它是否按规则工作。
素材来源(由使用者自己整理成文字,不要求学生上传文件):
· 一道真实错题的文字描述(题干 + 学生答案 + 正确答案 + 学生自述思路)
· 一段真实的对话片段(3-5 轮)
· 一个真实的知识点名称与所在章节
⚠️ 素材边界:
- 只用完成当前任务所需的最小信息;不收集真实姓名、学校班级全称、
联系方式、证件号、住址、家庭事件
- 用概括性描述替代精确身份信息
- 素材默认只在本次验证中使用,不进入长期档案;要长期保留必须单独征得同意
- 不要求上传 PDF、试卷扫描件、成绩单等文件——
收集文件带来不必要的隐私风险,对验证规则也没有帮助不要问"你记住我了吗"——直接用真实的问题测试。
好的测试方式:
拿 Step 3 准备的素材跑一遍,逐条核对:
① 是否先追问解题过程,而不是直接给答案
② 通用四维定位是否用了 shared/vocab.md 的取值
③ 轮次是否落在写好的预算内,两轮无回复时是否收尾
④ 语气与输出结构是否符合输出层的设定
⑤ 涉及情绪时,危机例外是否先于熔断触发
任一条不符合预期,记录哪里不对,进入 Step 5。每个SKILL至少需要2-3轮迭代才能稳定。
常见问题和修复方向:
① 追问语气太呆板
→ 修改输出层:写明语气与句式偏好,给两三个正例
② 总是直接给答案,没走提示阶梯
→ 修改规则层:写明默认最高级(L4/L5/L6)与升级条件
③ 记录的信息不够用
→ 修改记忆层:对照 dna-profile.schema.json 补字段,不自造字段名
④ 输出结构不稳定
→ 修改输出层:把结构、轮次预算、每步产出写成硬要求
💡 迭代规律:先把四层都写完,再逐轮测试优化。
不要未经测试就大量使用,先用5-10次对话热身。| 素材类型 | 具体形式 | 用来验证什么 |
|---|---|---|
| 一道真实错题 | 文字:题干 + 学生答案 + 正确答案 + 学生自述思路 | 四维定位、提示阶梯、轮次预算是否按规则走 |
| 一段真实对话 | 文字:3-5 轮的真实交互片段 | 语气、追问方式、退出规则是否生效 |
| 一个知识点 | 名称 + 所在章节 + 学段 | 学段判断、越纲标注、词表是否用对 |
纪律①:素材要真实但要脱敏
真实题目、真实思路 → 保留
真实姓名、学校班级、联系方式、家庭事件 → 一律不写进素材
纪律②:不要求上传文件
文字描述就足够验证规则是否生效。
不要在 SKILL 里写“请上传你的错题集 PDF”“发一份成绩单过来”这类要求。
纪律③:素材不等于长期档案
验证用的素材默认只在当次会话使用;
要长期保留必须单独征得同意,并按 shared/vocab.md §8 记录授权主体。教辅原题、历年真题一律 copyrightStatus = 仅存索引(shared/vocab.md §11):
只记“哪本书第几章第几题”,不整段复制原文。
自己改编的题标“改编”,公开可引用的标“公开可引用”。| 实践 | 核心规则 | 关键提示 |
|---|---|---|
| ① description 写清边界 | 3-6 句带学科词的触发语 + 明确“不处理什么、转给谁” | 触发冲突大多是 description 写太宽造成的 |
| ② 词表只引用不定义 | 需要错因、状态、掌握度、置信度时,写“见 shared/vocab.md §N” | 各自造词是全库不一致的头号来源 |
| ③ 字段名对着 schema 抄 | 接口章节只写 dna-profile / handover 里真实存在的字段 | 虚构 camelCase 路径会被 CI 拦下 |
| ④ 交接只用已有类型 | 七种 handoverType 之外的一律不发明 | 新类型要先改 schema,不能在正文里假设 |
| ⑤ 每步写轮次预算 | “① 收信息(1 轮)② 定位(1 轮)……全流程 ≤ 6 轮” | 没有预算的流程会拖成十几轮 |
| ⑥ 给长流程配快速模式 | 一个动作 ≤ 2 轮能走完的简版 | 时间紧时学生会直接放弃长流程 |
| ⑦ 数字要有出处 | 经验参数写“经验值,约…”并允许用户调整 | 无出处的比例和倍数一律删 |
| ⑧ 高中内容要标注 | 越出义务教育课标的术语同行标 ⚠高中,或放“初高衔接”小节 | CI 会按关键词扫这一项 |
写完一个 SKILL 后,逐条自检:
✅ 指标① 触发不冲突
description 里写明了不处理什么、转给谁;和相邻 SKILL 的触发语没有重叠
✅ 指标② 词表零自造
全文没有自定义的错因/状态/掌握度/置信度枚举,全部引用 shared/vocab.md
✅ 指标③ 字段可落地
接口章节里出现的每个字段名,都能在 dna-profile 或 handover schema 里找到
✅ 指标④ 边界齐全
技术边界一行 + 控制入口六行 +(涉情绪时)危机例外三行,都在
✅ 指标⑤ 流程有预算有出口
每步写了轮次,长流程有快速模式,有“两轮无回复即收尾”的退出
✅ 指标⑥ 学段说得清
frontmatter 的 grade_bands 与正文的适配说明一致;越纲内容有标注自检不过时的处理:
指标①不过 → 重写 description:删营销句,补“不处理什么、转给谁”。
指标②不过 → 把自造的词替换成 vocab 的取值;确实缺词就先改 vocab。
指标③不过 → 打开 schema 逐个核对;查不到的字段要么删,要么先加 schema。
指标④不过 → 从 shared/platform-conventions.md 与 shared/crisis-exception.md 复制标准段落。
指标⑤不过 → 给每一步加“(N 轮)”,并补一个 ≤2 轮的快速模式。
指标⑥不过 → 对照 shared/grade-bands.md 第四节确认适用性,不适用就直接写“不适用”。| 误区 | 表现 | 根本原因 | 修复方法 |
|---|---|---|---|
| 角色太宽 | 什么都管,什么都不精,和别的 SKILL 抢触发 | 角色层没限定职责 | 写出“只负责XX,不负责YY,YY 转给 zzz” |
| 规则太少 | 行为不稳定,时好时差 | 规则层只有 1-2 条 | 补到至少 4-6 条,含边界规则与退出规则 |
| 记忆太模糊 | 记了但不知道记了什么,也写不进 schema | 记忆层没有具体字段 | 改用“对于每个X,记录:[schema 里的字段名]” |
| 自造词表 | 同一个概念全库三种叫法 | 觉得自己的说法更顺口 | 一律引用 shared/vocab.md,需要新词就先改 vocab |
| 绝对禁令 | 写“永远不给答案”,把学生困死 | 把“不代做”误解成“不能讲” | 改写为提示阶梯 + 写明默认最高级 |
| 越权承诺 | 自己说“我明天提醒你”、自己按月出月报 | 没区分“入队”和“发送” | 提醒走 reminder_enqueue;周期报告改成用户请求后生成 |
SKILL 编写工具(开发者工具,不参与学生端运行时数据流)
──→ 产出:一份新的 SKILL.md 文本
←── 依据:shared/ 下的六份共享规范 + 两份 schema
⚠️ 本工具不读取任何学生档案、不发起交接、不写入 DNA。
它不在 handover-protocol 的 sender / recipient 枚举里,这是有意的。references/skill-templates-library.md — 七个场景的 SKILL 提示词模板(每个模板头部固定带危机例外片段)SECURITY_BASELINE.md — 仓库级 SKILL 安全与隐私基线shared/vocab.md — 全库唯一词表与阈值shared/platform-conventions.md — 能力代号、降级路径、控制入口、提醒入队契约shared/crisis-exception.md — 危机例外三行片段shared/hint-ladder.md — 提示阶梯与默认最高级对照表shared/ai-item-check.md — AI 出题自检协议shared/grade-bands.md — 学段参数表shared/dna-profile.schema.json — 档案字段定义shared/handover-protocol.schema.json — 七种交接类型💡 写在最后: 一个 SKILL 好不好用,不取决于它写得多长, 而取决于它的边界有多清楚—— 它知道自己做什么、不做什么、什么时候该把人交给别人。
© qizhitang, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
SKILL.md and 10 other files (references) in tools/xiaozhi-skill-creator of qizhitang/xiaozhi-skills.
Open the folder on GitHubat commit c65f2d4
Xiaozhi Skill Creator next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Xiaozhi Skill Creator this skillqizhitang/xiaozhi-skills | 105 | — | ~2.4k | Automated safety check: Pass | MIT | |
| Skill CreatorAzure/azqr | 796 | 89 repos | ~8.2k | Automated safety check: Pass | Apache-2.0 | |
| Claude Code Skill Developer Guidediet103/claude-code-infrastructure-showcase | 10k | 11 repos | ~3.5k | Automated safety check: Pass | MIT | |
| Darwin Skill Optimizeralchaincyf/darwin-skill | 6.2k | 1 repos | ~4.7k | Automated safety check: Pass | MIT | |
| Claude Code Command Developmentanthropics/claude-plugins-official | 38k | 10 repos | ~4.8k | Automated safety check: Pass | Apache-2.0 | |
| Claude Code Plugin Structureanthropics/claude-plugins-official | 38k | 10 repos | ~3.4k | Automated safety check: Pass | Apache-2.0 |
Azure/azqr
Create new skills, modify and improve existing skills, and measure skill performance.
diet103/claude-code-infrastructure-showcase
A guide to creating and managing Claude Code skills with auto-activation: skill-rules.json triggers, hooks, enforcement levels, YAML frontmatter and progressive disclosure.
alchaincyf/darwin-skill
Scores SKILL.md files on a nine-dimension rubric, then improves them in a keep-or-revert loop with independent judge agents, test prompts, git history and human checkpoints.
anthropics/claude-plugins-official
Explains how to write Claude Code slash commands: Markdown files with YAML frontmatter, arguments, file references, bash context and interactive prompts.
anthropics/claude-plugins-official
Explains the directory layout, plugin.json manifest and component organization of a Claude Code plugin, including auto-discovery and portable paths.
rohitg00/ai-engineering-from-scratch
Evaluates an Agent Skill bundle before release for structure, trigger quality, artifact improvement, script correctness, safety, installed-tree integrity and host portability.
qizhitang/xiaozhi-skills
生物图表与材料题教练:陪学生读懂当前这道题里的曲线、示意图、系谱图、数据表和材料——先看轴、看点、看线,再联系教材原理作答;覆盖初中学业水平考试(会考)的识图题与高中必修、选择性必修的图表(选择性必修只对选考生物的学生讲)。
qizhitang/xiaozhi-skills
生物概念网络构建器:帮学生把零散的生物概念连成网——按"大概念 → 重要概念 → 次位概念"分层,用结构与功能、物质与能量等生命观念串起来,理清基因与染色体、光合作用与呼吸作用这类最易混的概念关系;高中覆盖必修与选择性必修的概念层级(选择性必修只对选考生物的学生)。
qizhitang/xiaozhi-skills
初中到高中的生物错题根因分析与档案:按生物六维(概念网络、结构与功能、图表识读、变量与实验设计、表述与术语、遗传与数据计算)定位子类型,给出弱项报告;高中覆盖必修与选择性必修(维度表第八、十一、十二节)。
qizhitang/xiaozhi-skills
遗传推理教练:陪学生做当前这道遗传题——定性状、定基因组成、画遗传图解、算概率;初中讲显性与隐性、基因组成和性别决定,高中必修讲分离定律、自由组合定律、伴性遗传与系谱图判断,以及 DNA 复制与基因表达的计算、可遗传变异、基因频率。
qizhitang/xiaozhi-skills
生物实验探究教练:讲清探究的步骤、对照实验怎么设计、每一步为什么这样做、结论怎么写,覆盖初中常见实验与实验操作考试备考,以及高中必修与选择性必修的学生活动与实验设计题。
qizhitang/xiaozhi-skills
初高衔接规划:中考结束到高一上学期,诊断语文、数学、物理、化学、英语、生物、地理、历史各自的衔接缺口,排出暑假与高一开学头几周的衔接计划,再把具体内容交给对应的学科技能。
Categories
写一个新 SKILL 时用的编写工具:四层结构(角色/规则/记忆/输出)、安全与隐私边界、五步落地流程、常见问题诊断. An agent skill from qizhitang/xiaozhi-skills. Xiaozhi Skill Creator is an agent skill from qizhitang/xiaozhi-skills.
Xiaozhi Skill Creator fits situations like: tasks that involve Skill authoring.
Run `npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a claude-code`. Or copy the skill folder (tools/xiaozhi-skill-creator in qizhitang/xiaozhi-skills) into .claude/skills/xiaozhi-skill-creator in your project. Claude Code loads it when a task matches its description.
Run `npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a codex`. Or copy the skill folder (tools/xiaozhi-skill-creator in qizhitang/xiaozhi-skills) into .agents/skills/xiaozhi-skill-creator in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add qizhitang/xiaozhi-skills --skill xiaozhi-skill-creator -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/xiaozhi-skill-creator, .gemini/skills/xiaozhi-skill-creator, .github/skills/xiaozhi-skill-creator and .opencode/skills/xiaozhi-skill-creator in your project.
SKILL.md names no scripts, command-line tools or credentials: Xiaozhi Skill Creator is instructions for the agent only. Compatibility (from SKILL.md): WorkBuddy / SkillHub / OpenClaw / ClawHub.
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
Xiaozhi Skill Creator is published under the MIT licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.
About 2.4k tokens (SKILL.md is roughly 9.7k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 3.6k tokens, read only when the agent opens those files.
Skills that share tags, products or a category with Xiaozhi Skill Creator: Skill Creator (Azure/azqr, 796 stars), Claude Code Skill Developer Guide (diet103/claude-code-infrastructure-showcase, 10k stars), Darwin Skill Optimizer (alchaincyf/darwin-skill, 6.2k stars) and Claude Code Command Development (anthropics/claude-plugins-official, 38k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
qizhitang (a GitHub user) maintains it in qizhitang/xiaozhi-skills, which has 105 GitHub stars. The repository holds 84 skills in this directory. The repository was last updated on October 9, 2026.
Source: qizhitang/xiaozhi-skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.