Workflows2026年5月13日·1 分钟阅读

Agent Governance Toolkit — Policy Guardrails for Agents

Microsoft's Agent Governance Toolkit adds policy checks, red-team scans, evidence verification, and runtime guardrails to autonomous agents.

Agent Toolkit
Agent Toolkit · Community
Agent 就绪

这个资产可以被 Agent 直接读取和安装

TokRepo 同时提供通用 CLI 命令、安装契约、metadata JSON、按适配器生成的安装计划和原始内容链接，方便 Agent 判断适配度、风险和下一步动作。

Native · 94/100策略：允许
Agent 入口
任意 MCP/CLI Agent
类型
Workflow
安装
Pip
信任
信任等级：Established
入口
pip install agent-governance-toolkit[full]
通用 CLI 安装命令
npx tokrepo install bfd4f4dc-d93b-558a-850f-6e3c49c99cb7
安装契约元数据 JSON适配计划原始内容
介绍

微软的 Agent Governance Toolkit 为自主 agent 系统加入策略校验、红队扫描、证据验证与 OWASP 风格运行时护栏，适合安全要求较高的场景。

Best for: 需要围绕自主 agent 增加策略执行、运行证据与红队检查的安全敏感型团队

Works with: Python 安装、CI 流水线、agent framework、证据文件、安全审查、自主执行护栏

Setup time: 8-12 minutes

Key facts (verified)

  • GitHub：1518 stars · 102 forks；最近更新 2026-05-13。
  • 许可证：MIT；作者头像与仓库链接均已通过 GitHub API 复核。
  • README 中核对过的入口命令：pip install agent-governance-toolkit[full]。

Main

这个工具箱面向的是“不能只靠信任 agent” 的环境，它把治理要求变成可执行检查，而不是停留在口头规范。

命令集合本身就很有操作意义：agt doctor 验证安装，agt verify 检查策略与证据，agt red-team scan 则把安全评估前移到开发流程中。

README 直接对齐 10/10 OWASP Agentic Top 10 也是重要信号：对需要讲清安全故事的团队来说，它不仅有控制点，也有对外可解释性。

Source-backed notes

  • README starts with a 90-second setup flow using pip install agent-governance-toolkit[full].
  • It exposes agt doctor, agt verify, agt verify --strict, and agt red-team scan commands.
  • The project explicitly claims coverage for 10/10 items in the OWASP Agentic Top 10 framing.

FAQ

问：Is this only documentation? 答：No. The project ships executable commands for doctor, verify, strict verification, and red-team scans.

问：Why use it in CI? 答：It can fail weak evidence or risky patterns before autonomous behavior reaches production.

问：Who benefits most? 答：Teams with compliance, security, or governance obligations around AI agents.

🙏

来源与感谢

Source: https://github.com/microsoft/agent-governance-toolkit > License: MIT > GitHub stars: 1518 · forks: 102

讨论

登录后参与讨论。
还没有评论，来写第一条吧。

相关资产

Agent Safehouse — Sandbox macOS Coding Agents

Agent Safehouse sandboxes local coding agents on macOS using sandbox-exec deny-first profiles, limiting access to only approved files and integrations.

Workflows
Agent Toolkit

Obsidian Agent Client — Bring Agents into Obsidian

Obsidian Agent Client adds ACP-based agent integrations (Claude Code/Codex/Gemini) to Obsidian; install via BRAT or Releases and use adapters (2,016★).

Workflows
Agent Toolkit

beava — Live Features for Agent Guardrails

beava turns live product events into guardrail-ready agent features; verified 130 stars, Rust core, and a 60-second quickstart for teams skipping Kafka.

Workflows
Agent Toolkit

Stripe Agent Toolkit — Payments SDK for LLM Agents

Stripe Agent Toolkit gives LangChain, OpenAI Agents SDK, CrewAI, and Vercel AI SDK direct tool access to charges, customers, and refunds.

Scripts
Stripe
◈首页🔍搜索👤我的