不是翻译 —— 多数文档描述的行为已经不存在了。 先修一个更基本的问题:我们不拥有 u-claw.org 域名,那是上游的。所以之前写在 README、诊断包提示、联系方式里的 help@u-claw.org 全都会把用户的问题发给 上游 —— 一个没有理由回复的人。改为指向我们自己的 issue tracker,并加断言 禁止再出现指向该域名的支持入口。 重写(内容过时,不是语言问题): - install/README.md —— 还写着 10 个中国技能、DeepSeek 优先、国内镜像。 现在按实际流程写:技能读 manifest、模型菜单 Gemini 优先、bundle 有 SHA256 校验。并如实写明 curl|bash 在受管企业电脑上会被 EDR 拦。 - CLAUDE.md 的模型配置整节 —— 还在描述虾盘云首选卡片和 12 个 provider, 那个界面已经换成单框 Key 输入了。 - SECURITY.md —— 安全报告原本指向上游维护者个人邮箱。fork 之后那条路由 是错的:漏洞会发给写不了这份代码、也修不了的人。 - CONTRIBUTING.md —— 补上 fork 关系、pre-push 钩子怎么装、以及 `node --test tests/` 为什么不能用。 翻译并保留: - bootable/README.md、TROUBLESHOOTING.md —— 面向用户,顺带把 「国内镜像」「小米/华为 BIOS 按键」等换成目标市场的实际情况 HANDOFF.md 重写为一份事故复盘:原文一半是过时的一次性交接笔记(引用的 website/guide.html 已不在本仓库),另一半是 persistence.dat 未格式化导致 启动失败的排查记录 —— 后者有长期价值,尤其是「读 offset 1080 的两字节 验证 ext4」这个判断方法,已同时写进 bootable/README.md。 bootable/IMPROVEMENTS_SUMMARY.md 保留中文,加了说明:它是上游 fork 前的 历史改进记录,没人引用,描述的是已完成的工作而非当前行为。翻译它反而会 让人误以为是现行文档。 新增 skills/en/uclaw-help —— 把「怎么用、东西在哪、出问题怎么办」做成 内置知识,每个角色都装。方案 C10.8:能问的产品才是不需要学的产品。
2.8 KiB
Security policy
Reporting a vulnerability
Please do not put details in a public issue.
Open a report at https://gitea.fanghe.it.com/zhenghy/u-claw/issues — mark it clearly as a security report and keep the details minimal until we can arrange a private channel.
This is a fork. Upstream is
github.com/dongsheng123132/u-claw. If the issue is in code we inherited unchanged, it affects their users too — please tell them as well. If it is in something this fork changed (the launcher, the i18n layer, the skill installer, the key-check endpoint, the self-heal routines), it is ours.We do not yet have a dedicated security address of our own.
origin.jsonrecords that gap; until it is filled, the issue tracker is the route.
We will acknowledge within 3 working days and give a first assessment within 14 days. Please give us a chance to fix it before disclosing publicly.
What helps
- Which module —
portable/,install/,bootable/ - Version or commit hash
- How it is triggered, and from where: local, LAN, or the internet
- Steps to reproduce, or a proof of concept
- What you think the fix looks like, if you have a view
What is in scope
- Command injection, path traversal, arbitrary write or code execution in U-Claw's own scripts
- Injection through the loading or writing of
data/.openclaw/openclaw.json - Launchers (
START HERE - *,lib/start.mjs) being hijacked through a malicious directory name or environment variable - Man-in-the-middle exposure on the one-line install path
(
curl | bash,irm | iex) - Privilege escalation in what the bootable USB scripts produce
- The key-check endpoint (
/api/test-key) — it handles API keys server-side - The diagnostics report leaking a secret it should have redacted
What is not
- Upstream dependencies — Node.js, OpenClaw, Electron, Ventoy, Ubuntu. Report those to their projects; we track and upgrade versions.
- A user publishing their own API key in a screenshot or a public repo.
- Physical access to the drive. If someone has the drive, they have the data — that is the trade a portable product makes, and it is stated plainly in the README rather than defended against.
- What an attacker can do once they already have admin or root on the machine.
Supported
| Module | Status |
|---|---|
portable/ |
current main — reports accepted |
install/ |
current main — reports accepted |
bootable/ |
current main — reports accepted |
u-claw-app/ |
deprecated 2026-06-19, no longer built or published |
| older tags | severity assessed, but no backported patches |
There are no stable version numbers yet; fixes go straight to main.
Credit
Anyone who reports responsibly is credited in the fixing commit and the release notes, unless you would rather not be.