安全修复优先针对当前 main 分支和最新发布版本。旧版本是否修复,取决于问题影响、修复成本和维护者判断。
Security fixes primarily target the current main branch and the latest release. Whether older versions are fixed depends on impact, maintenance cost, and maintainer judgment.
| 版本 / Version | 安全支持 / Supported |
|---|---|
当前 main / Current main |
✅ |
| 最新发布版 / Latest release | ✅ |
| 更早版本 / Older versions | 视情况而定 / Case by case |
请不要通过公开 Issue、Pull Request、讨论区或公开群消息报告安全漏洞。请优先使用仓库 GitHub 页面上的私密漏洞报告入口(Security → Report a vulnerability,若该功能已启用)。
Do not report security vulnerabilities through public Issues, Pull Requests, discussions, or public group messages. Prefer the private vulnerability reporting entry on the repository’s GitHub page (Security → Report a vulnerability, if enabled).
如果私密漏洞报告入口不可用,请加入开发者 QQ 群 263402786,并私信 Eternal-Wanderer-Vegetable,说明你希望私下报告安全问题。请不要发送真实用户数据、API 密钥或完整数据库;只提供验证问题所需的最小脱敏信息。
If private vulnerability reporting is unavailable, join the developer QQ group 263402786 and privately contact Eternal-Wanderer-Vegetable, stating that you want to report a security issue privately. Do not send real user data, API keys, or complete databases; provide only the minimum redacted information needed to verify the issue.
报告内容建议包括:
- 受影响的版本、提交或配置。
- 漏洞类型、影响范围和潜在后果。
- 最小复现步骤或概念验证。
- 你认为可行的缓解或修复建议。
- 是否已经向其他人披露,以及期望的联系渠道。
Include the affected version, commit, or configuration; vulnerability type, scope, and potential impact; minimal reproduction steps or proof of concept; possible mitigations; and any prior disclosure or preferred contact channel.
维护者会尽快确认收到报告,并在完成初步评估后告知后续处理方式。确认漏洞后,维护者将根据情况修复、发布安全更新、在发布说明中致谢报告者(除非对方要求匿名),并在适当时公开漏洞摘要。
The maintainer will acknowledge receipt as soon as practical and explain the next steps after an initial assessment. Once confirmed, the maintainer may fix the issue, publish a security update, credit the reporter in release notes unless anonymity is requested, and publish a vulnerability summary when appropriate.
Stella 可以连接在线模型服务、OneBot 实现和第三方插件。请在报告中明确说明数据是否会离开本机;在线服务的请求内容取决于你的部署配置。任何公开示例都必须去除聊天内容、个人信息、密钥和其他凭据。
Stella can connect to online model services, OneBot implementations, and third-party plugins. When reporting an issue, state whether data leaves the local machine; requests to online services depend on your deployment configuration. Remove chat content, personal information, keys, and other credentials from all public examples.