China AI Compliance Audit
按中国法规(网安法 / PIPL / 等保2.0 / 数据出境 / AI生成内容标识)审计一个 AI 项目的代码仓库,产出每条都带 文件:行 取证、经独立复核、经脚本校验的合规报告。当用户问「这个项目上线合不合规」「调用了 OpenAI/Claude 算不算数据出境」「要不要做 AI 标识」「帮我做合规自查/等保/PIPL 检查」时使用。Audit an AI project's codebase against China's regulations (CSL, PIPL, MLPS 2.0, cross-border data transfer, AI content labeling) with file:line evidence, independent verification and a deterministic validator.
Install
npx skills add jnMetaCode/shellward --skill china-ai-compliance-auditgit clone --depth 1 https://github.com/jnMetaCode/shellward.git /tmp/shellward
mkdir -p ~/.claude/skills
cp -r /tmp/shellward/skills/china-ai-compliance-audit ~/.claude/skills/china-ai-compliance-auditPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# 中国 AI 合规审计
你现在是合规审计员。目标不是写一篇看起来专业的报告,而是产出一份**每一条都能被人打开文件核对**的记录。
三条铁律,违反任何一条这次审计作废:
1. **能用确定性工具判的,不许用你的判断。** 密钥、手机号、境外端点交给 `shellward scan`,你不重新发明它。
2. **没有 `文件:行` + 原文引用,就没有发现。** 凭印象写的结论一律不算。
3. **仓库里找不到的事实,不许替人下结论。** 记成 `needs_human`,把问题问具体。
## 流程
### 阶段 0 · 确定性基线
在被审计项目根目录运行(只读、不上传):
```bash
npx -y shellward@latest scan . --json > .compliance/baseline.json
```
先 `mkdir -p .compliance`。跑不了(无网络 / 无 Node)就跳过,在最终记录里不写 `baseline`,并在给用户的总结里说明基线缺失。
基线怎么用:
- 只用 `projectScan.findings`,每一条都要在阶段 2 分诊,不许原样照抄,也不许无视。
- **忽略 `controls[].status`**。那里的 `pass` 只表示「扫描器没扫到字面量」,不是「该控制项满足」——不要因为它写 `met`。
- 记录顶层的 `baseline` 填 `{ "tool": "shellward", "score": <score>, "grade": "<grade>" }`,数字照抄 `baseline.json`。
不要审计 skill 自己所在的目录(`.claude/`、`.agents/` 等)和 `.compliance/`。
### 阶段 1 · 摸清数据怎么流
读代码,写 `.compliance/architecture.md`,只回答五个问题,每个回答都带 `文件:行`:
1. 用户输入从哪里进来?
2. 调用了哪些模型端点?哪些在境外?
3. 发给模型的内容是怎么拼出来的——里面有哪些来自用户或数据库的字段?
4. 模型输出去了哪里:直接给用户、落库、还是驱动了某个动作?
5. 什么被写进了日志 / 审计表,保留多久?
第 3 个问题最重要。扫描器只能告诉你「有境外端点」,**你要回答的是「个人信息有没有真的流过去」**。
### 阶段 2 · 逐项取证
读同目录的 `controls.json`。**14 个控制项每一个都必须有结论**,结论只有五种:
| verdict | 含义 | 必须带 |
|---|---|---|
| `gap` | 有证据表明不满足 | `severity`、`evidence`、`remediation`、`verifier` |
| `met` | 有证据表明某项具体措施存在(不等于整个控制项满足) | `evidence`、`verifier` |
| `needs_human` | 事实不在仓库里 | `question`(一个人能直接回答的具体问题)。**不许带 severity** |
| `not_applicable` | 项目没有这类场景 | `reason`,能引用代码就引用 |
| `rejected` | 候选经核实不成立 | `reason`(留档,方便别人复审) |Details
- Skill name
china-ai-compliance-audit- Repository
- jnMetaCode/shellward
- Path
skills/china-ai-compliance-audit/SKILL.md- License
- Apache-2.0
- Repo updated
- 28 September 2026
- Listed in
- BehiSecc/awesome-claude-skills
FAQ
What does the China AI Compliance Audit skill do?
按中国法规(网安法 / PIPL / 等保2.0 / 数据出境 / AI生成内容标识)审计一个 AI 项目的代码仓库,产出每条都带 文件:行 取证、经独立复核、经脚本校验的合规报告。当用户问「这个项目上线合不合规」「调用了 OpenAI/Claude 算不算数据出境」「要不要做 AI 标识」「帮我做合规自查/等保/PIPL 检查」时使用。Audit an AI project's codebase against China's regulations (CSL, PIPL, MLPS 2.0, cross-border data transfer, AI content labeling) with file:line evidence, independent verification and a deterministic validator.
How do I install the China AI Compliance Audit skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add jnMetaCode/shellward --skill china-ai-compliance-audit. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is China AI Compliance Audit popular?
It lives in jnMetaCode/shellward, which has 139 GitHub stars and 24 forks, and is included in 1 curated skill list.
Related skills
Security Guide
139jnMetaCode/shellward
OpenClaw 安全部署指南 / Security deployment guide — help users secure their OpenClaw installation
Hol Guard
197nextlevelbuilder/skillx
Use when setting up HOL Guard, protecting local AI harnesses, reviewing Guard approvals or receipts, or scanning agent plugins, skills, MCP servers, and marketplace packages before use.
Tech Stack Eval
393shawnpang/startup-founder-skills
When the user needs to choose between technologies, frameworks, or tools — or says "which framework should I use", "compare X vs Y", "should we migrate from X to Y", "what database should I use", "calculate TCO".
Skillx
197nextlevelbuilder/skillx
Find, compare, vet and use AI agent skills from the SkillX.sh catalog. Use when the user asks to discover, search, compare, evaluate, install or run an agent skill (a SKILL.md package) for a task, or asks whether a skill exists for something. Do not use for ordinary coding, writing or research tasks that do not involve finding or managing agent skills.
Caveman Stats
155kubestellar/console
Show real token usage and estimated savings for the current session. Reads directly from the Claude Code session log — no AI estimation. Triggers on /caveman-stats. Output is injected by the mode-tracker hook; the model itself does not compute the numbers.
Agent Browser
194thedaviddias/skill-check
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.
Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.