Security audit
Security checks for vulnerabilities and agentic risk
This is an LGD onboarding guide with optional installation steps, broad activation phrases, and outdated verification records.
Use this skill for LGD-specific onboarding. Generic activation should not authorize installing the wider collection. Before following installation commands, review the selected skills and pin the installer and source revision where possible. Prefer individual installation if you do not want global changes. The bundled integrity records do not match the current skill file, and this review does not cover externally downloaded components.
The installation command executes an unpinned npm CLI and installs skills globally from a repository without an immutable revision. Although this is not an MCP server declaration, the underlying supply-chain risk is real: compromised upstream releases or repository changes could execute code with the user's permissions or introduce persistent agent instructions.
This onboarding table repeats the recommendation to use an unversioned npx installer, reinforcing the mutable installation path described earlier. It is an abbreviated installation recommendation rather than an MCP configuration, but following it still exposes users to upstream package changes.
The skill explicitly requires activation for generic onboarding and governance questions that do not necessarily request LGD. This can redirect unrelated user tasks into an author-specific framework and installation workflow, broadening the reach of its instructions beyond the user's intended scope.
The index lists generic phrases such as 'onboarding' and 'getting started' as standalone activation conditions without requiring LGD context. A router relying on this index could load unrelated framework instructions and steer the user's task toward the author's ecosystem.
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
- 说明：该文件为发布流程自身工具，不在 zip 包内（白名单打包已排除），不属被审计交付物。
## 📋 详细检查结果
- **命令执行与权限检查**: 包内 0 次命中（curl|bash / eval / exec / subprocess / os.system / sudo / nohup 均无）。SKILL.md 中 `npx skills add` 与 `git clone` 为面向读者的安装指引，需用户手动复制执行，属「示例/说明」——按审计边界不属于自动执行，不报告。
- **文件操作与敏感路径检查**: 0 次命中（~/.ssh、.aws、.env、C:\Users 等均无）。
- **网络请求检查**: 19 个 URL 全部为官方/自有域名（github.com/zhaoxinghua09-cell/*、github.com/Medxpert-org/lgd-engine、github.com/LGD-Eval/skill-audit、medxpert.cn/*、doi.org、market.lobehub.com）；全部经 2026-11-04 实测 live；无 Base64 可疑载荷。
- **远程脚本深度分析**: 不适用——包内无任何自动下载+执行的远程脚本。
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
- 说明：该文件为发布流程自身工具，不在 zip 包内（白名单打包已排除），不属被审计交付物。
## 📋 详细检查结果
- **命令执行与权限检查**: 包内 0 次命中（curl|bash / eval / exec / subprocess / os.system / sudo / nohup 均无）。SKILL.md 中 `npx skills add` 与 `git clone` 为面向读者的安装指引，需用户手动复制执行，属「示例/说明」——按审计边界不属于自动执行，不报告。
- **文件操作与敏感路径检查**: 0 次命中（~/.ssh、.aws、.env、C:\Users 等均无）。
- **网络请求检查**: 19 个 URL 全部为官方/自有域名（github.com/zhaoxinghua09-cell/*、github.com/Medxpert-org/lgd-engine、github.com/LGD-Eval/skill-audit、medxpert.cn/*、doi.org、market.lobehub.com）；全部经 2026-11-04 实测 live；无 Base64 可疑载荷。
- **远程脚本深度分析**: 不适用——包内无任何自动下载+执行的远程脚本。
npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.