Enhance security policy with detailed reporting and scope for GHSA

[noyp-lasso]

Expanded the security policy to include detailed reporting guidelines, response targets, and scope of vulnerabilities. Added sections on coordinated disclosure and safe harbor for researchers.

变更描述

🔥 务必注意，本项目不允许提交未经 “Human Review” 过的代码，不允许 Agent 未经人类审阅，直接提交 PR。不允许 Agent 直接提交回复。

Code is Cheap, Show Me Your Thoughts!

简要描述这个 PR 做了什么

变更类型

• 新功能
• Bug 修复
• 文档更新
• 其他

测试

• 已在 Docker 环境测试
• 相关功能正常工作

相关日志或者截图

说明

（可选）有什么需要特别说明的吗？

---

💡 提示: 提交前可以运行 make lint 和 make format 检查代码规范

[gemini-code-assist]

Code Review

This pull request introduces a new SECURITY.md file to establish the project's security policy, detailing vulnerability reporting, disclosure timelines, and scope. The review feedback suggests improving the accuracy and consistency of the document by renaming 'Yuxi-Know' to 'Yuxi' to match the project's branding, and updating the reference from 'frontend' to 'web' to align with the actual repository directory structure.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

项目中前端目录的实际名称为 web（如 CONTRIBUTING.md 中所述），而非 frontend。为了与仓库的实际目录结构保持一致，建议将 frontend 修改为 web，以避免混淆。

[xerrors]

Thanks for the contribution.

Before we proceed with this PR, please clarify whether the accompanying security analysis was performed against the latest main branch. Several claims appear to be based on older code paths or deployment assumptions that may no longer match the current repository.

Please confirm, through the private disclosure channel:

• The exact Yuxi commit SHA that was reviewed
• The date the repository was cloned or analyzed
• Whether the finding was reproduced against the current main branch
• The exact DeepAgents version and dependency lockfile used
• Which impact claims were directly validated and which were inferred from source or deployment configuration

Please do not post any vulnerability details, affected code paths, payloads, reproduction steps, or impact analysis in this public PR.

For this PR itself, please also update frontend to web and use Yuxi consistently instead of Yuxi-Know.

[noyp-lasso]

1. Exact Yuxi commit SHA reviewed

   • 2c8ff10dc6bc (branch main, committed 2026-07-08 10:20:56 +0800).
   • This is also the CURRENT main HEAD as of 2026-07-12 — main has not advanced since our
     original review, so we are not working from a stale tree.

2. Date the repository was cloned / analyzed

   • Original analysis: 2026-07-11 (our evidence artifacts are timestamped 2026-07-11T14:20Z).
   • Re-verification against current main: 2026-07-12 (this message).

3. Whether the finding was reproduced against the current main branch

   • Yes. On 2026-07-12 we re-confirmed every cited code path exists on current main
     (2c8ff10dc6bc);

4. Exact DeepAgents version and dependency lockfile used

   • Your lockfile pins DeepAgents 0.6.7 exactly: backend/uv.lock (name = "deepagents",
     version = "0.6.7"; project spec deepagents>=0.6.7).
   • Our original harness ran the then-installed DeepAgents 0.6.12. To remove any
     version-drift doubt, on 2026-07-12 we re-ran our deterministic proof pinned to 0.6.7
     (uv run --with deepagents==0.6.7). Result on 0.6.7 is identical

5. Which impact claims were directly validated vs inferred
   We are not able to answer this one in the open pull request: a meaningful
   validated-vs-inferred breakdown necessarily describes the vulnerability's impact, the
   affected components, and the reachability/exploitation path — i.e. exactly the vulnerability
   details, affected code paths, and impact analysis you have (rightly) asked us not to post
   publicly. We cannot provide it here without exposing sensitive details of the issue.
   Please confirm the private disclosure channel (the private GitHub Security Advisory thread, or another channel you
   prefer) and we will provide the complete impact breakdown there immediately.

[xerrors]

Thanks for the reminder. The risk points mentioned in your email have been received, and we will address them in the near term.

Back to this PR: please handle this PR seriously. Before we continue reviewing it, please first address the PR-specific issues already mentioned above, including using Yuxi consistently instead of Yuxi-Know and updating frontend to web. Please also keep any vulnerability details, affected code paths, payloads, reproduction steps, or impact analysis out of this public PR and continue those details only through the private disclosure channel.

[noyp-lasso]

Thanks for the review and the clear guidance.

I've addressed the PR-specific items in the latest commit: the policy now uses "Yuxi" consistently (no more "Yuxi-Know") and refers to "web" instead of "frontend". This PR only adds the general SECURITY.md policy and contains no vulnerability details.

Understood on keeping all vulnerability details out of this public PR. I'll send the full private details once GitHub Private Vulnerability Reporting / GHSA is enabled (it currently appears off) — please enable it or open a draft advisory and add me as a reporter.