模板详情

安全 AI 编程基线

跨工具安全基线，适合希望 AI agent 谨慎处理凭据、迁移、网络请求和破坏性操作的团队。

CursorClaude CodeCodexWindsurfContinueAGENTS.mdCLAUDE.md.cursor/rules
913投票
21,090安装量
已验证状态

安装预览

命令仍是 MVP 预览。真实使用前请检查目标文件和覆盖行为。

npx rulesmarket add secure-ai-coding-baseline
返回模板库
AGENTS.md
# Secure AI Coding Baseline

- Treat credentials, tokens, cookies, private files and production data as sensitive.
- Never print, commit, upload or transmit secrets.
- Ask before destructive actions, permission changes, production writes or external submissions.
- Prefer allowlisted scripts and documented project commands.
- Review dependency changes for license, maintenance and security impact.
- Validate inputs at boundaries and keep authorization checks close to the protected action.
- When uncertain, produce a minimal patch and explain remaining risk instead of guessing.

这足够覆盖企业安全要求吗？

不够。它是基础指令文件，不是完整的安全制度或策略替代品。

每个仓库都应该使用吗？

它适合作为 AI 辅助仓库的起点，尤其是 agent 可以运行工具的仓库。