跳转到主要内容
此内容尚未提供您的语言版本,正在以英文显示。

Review Agent Governance

插件 已验证 活跃

Require a human approval signal before an AI agent can post PR reviews, comments, merges, or writes to CI config. Cedar-gated, receipt-signed, designed for the Hermes-style failure mode where a review bot posts without oversight.

1 个 Skill 0 个 MCP
目的

To prevent AI agents from making unauthorized or erroneous changes to code reviews, PRs, or CI configurations by requiring explicit human approval and providing a cryptographically verifiable audit trail.

功能

  • Mandatory human approval for review-surface actions
  • Cedar policy for declarative gating
  • Ed25519-signed receipts for auditability
  • Slash commands and flag files for approval windows
  • Automatic denial of unauthorized actions

使用场景

  • Ensuring human oversight on AI-generated PR reviews.
  • Preventing AI agents from merging code without explicit human sign-off.
  • Auditing all AI agent actions on code review surfaces.
  • Controlling AI agent modifications to CI/CD workflow files.

非目标

  • General policy enforcement for all tool calls (use `protect-mcp` for this).
  • Replacing the need for human review entirely.
  • Automating the approval process itself.

安装

请先添加 Marketplace

/plugin marketplace add wshobson/agents
/plugin install review-agent-governance@claude-code-workflows

质量评分

已验证
99 /100
13 days ago 分析

信任信号

最近提交15 days ago
星标35.3k
许可证MIT
状态
查看源代码