‹ 首页

agent-governance

@mahsumaktas · 收录于 1 周前 · 上游提交 4 个月前

Agent islemlerini policy'ye gore denetle. Hassas islem oncesi governance kontrolu calistir. Tetikleyiciler: destructive komut, dosya sistemi erisimi, rate limit, butce kontrolu, trust score yonetimi.

场景权限管控安全审计操作审批

适合你，如果需要在代理执行危险操作前进行策略审核

/ 下载安装
https://github.com/mahsumaktas/agent-evolution-kit/tree/main/skills/agent-governance ↗
/ 通过 npx 安装 校验哈希
npx oh-my-skill add mahsumaktas/agent-evolution-kit/agent-governance
/ 通过 bash 安装
curl -fsSL https://oh-my-skill.com/install.sh | bash -s -- mahsumaktas/agent-evolution-kit/agent-governance
/ 已经装过?验证本机副本,不用重装
npx oh-my-skill verify mahsumaktas/agent-evolution-kit/agent-governance
安装目标可用 --agent / --scope 或 --to 明确指定;省略时只会在唯一已存在的 agent 目录上自动选择,零命中或多命中会停止并提示。content_hash 缺失或不一致均拒装。
1GitHub stars
~644上下文体积 · 单文件
索引托管

怎么用

商店整理自技能原文 · 版本 c599c87 · 表述以原文为准
它做什么

装上后，Claude 会在执行敏感操作前检查策略文件，根据信任等级、文件系统权限、速率限制和预算等规则决定允许、警告或拒绝操作，并记录审计日志。

什么时候触发

当 Claude 要执行破坏性命令、访问文件系统、触发速率限制或涉及预算控制等敏感操作时，会触发治理检查。

装好后可以这样说
触发破坏性操作警告，记录日志。
执行信任分数查询。
查看审计记录。
技能原文 SKILL.md作者撰写 · MIT · c599c87

Agent Governance

Genel Bakis

Agent'larin islemlerini policy tabanli denetleme sistemi. Her hassas islem oncesi governance kontrolu yapar, audit log'a kaydeder.

Temel ilke: Guven ama dogrula. Her islem loglanir, hassas islemler policy'ye gore degerlendirilir.

Policy Dosyasi

Konum: ~/.agent-evolution/config/governance.yaml

Her agent icin tanimlanmis:

  • trust_level: 0-1000 arasi guven skoru
  • filesystem: allowed/denied dizin listeleri
  • rate_limits: Saatlik islem limitleri
  • budget: Gunluk maliyet limiti
Calisma Modlari

| Mod | Davranis | |-----|----------| | audit-only | Logla + uyar, engelleme (baslangic modu) | | blocking | Logla + engelle (2 hafta veri sonrasi) |

Kullanim
Islem Oncesi Kontrol

Her hassas islem oncesi:

exec: ~/.agent-evolution/scripts/oracle-governance.sh check <agent> <action> <args>

Ornekler:

exec: ~/.agent-evolution/scripts/oracle-governance.sh check primary-agent exec "ls -la ~/clawd"
# → ALLOW (allowed filesystem)

exec: ~/.agent-evolution/scripts/oracle-governance.sh check social-agent exec "cat ~/.ssh/id_rsa"
# → DENY (denied filesystem)

exec: ~/.agent-evolution/scripts/oracle-governance.sh check primary-agent exec "rm -rf /tmp/test"
# → WARN (destructive action)
Audit Log Goruntuleme
exec: ~/.agent-evolution/scripts/oracle-governance.sh audit --last 10
exec: ~/.agent-evolution/scripts/oracle-governance.sh audit primary-agent --last 5
Haftalik Rapor
exec: ~/.agent-evolution/scripts/oracle-governance.sh report --weekly
Trust Score Yonetimi
exec: ~/.agent-evolution/scripts/oracle-governance.sh trust primary-agent        # mevcut skoru goster
exec: ~/.agent-evolution/scripts/oracle-governance.sh trust social-agent +50      # skor artir
exec: ~/.agent-evolution/scripts/oracle-governance.sh trust finance-agent -100      # skor azalt
exec: ~/.agent-evolution/scripts/oracle-governance.sh trust tithonos =750   # skor ayarla
Istatistikler
exec: ~/.agent-evolution/scripts/oracle-governance.sh stats
Hassas Islem Kategorileri

| Kategori | Ornekler | Varsayilan Davranis | |----------|----------|---------------------| | Destructive | rm -rf, git push --force, drop table | WARN + log | | Financial | trade, transfer, payment | DENY + log | | External | curl POST, webhook, smtp | WARN + log | | Filesystem denied | ~/.ssh, ~/.aws erisimi | DENY + log | | Rate limit | exec_per_hour asildi | WARN + log |

Entegrasyon Kurallari
  1. Yeni agent olusturuldiginda: governance.yaml'a entry ekle
  2. Trust ihlali tespit edildiginde: trust score dusur + Discord alert
  3. 2 hafta audit-only sonrasi: Veriyi analiz et, blocking'e gec
  4. Haftalik: oracle-governance.sh report --weekly calistir
Iliskili Dosyalar
  • Policy: ~/.agent-evolution/config/governance.yaml
  • Script: ~/.agent-evolution/scripts/oracle-governance.sh
  • Audit DB: ~/.agent-evolution/governance/audit.db
  • Trust Scores: ~/.agent-evolution/governance/trust-scores.json
按 MIT 许可原样转载,未经改动 · 在 GitHub 查看 →

评论

登录即可评论;带「已验证安装」的,是发布者名下有本店的安装或持有记录。
…