agent-guardrails
bởi olmmlo-cmd
Coding Agents & IDEsTóm tắt
Stop AI agents from secretly bypassing your rules.
Mô tả
Kỹ năng này làm gì
Mechanical enforcement layer for AI agent project standards using git hooks, secret detection, deployment verification, and import registries. Born from real production incidents including server crashes, token leaks, and code rewrites. Enforces rules that markdown and prompt instructions cannot reliably maintain.
Git hooks and code-level checks enforce standards at 95-100% reliability compared to 40-70% for prompt rules and markdown instructions.
Khi nào nên dùng
- Blocking AI agents from reimplementing existing functions instead of importing them
- Catching hardcoded API keys before they reach version control
- Verifying deployed code matches written code after agent tasks
- Preventing deployment gaps where features are coded but never wired into production
- Propagating enforcement improvements across projects via skill update feedback loops
Quy trình ví dụ
User asks: set up guardrails on my Python project
- 1
Runs install.sh to install git pre-commit hook and scaffolding into the project
- 2
Copies registry-template.py to create a module import registry
- 3
Adapts agents-md-template.md into project AGENTS.md with mechanical enforcement rules
- 4
Runs create-deployment-check.sh to generate .deployment-check.sh and DEPLOYMENT-CHECKLIST.md
- 5
Installs the pre-commit-deployment git hook template
Project now blocks secret commits, detects duplicate implementations, and verifies deployment wiring before any commit is accepted
Cài đặt
clawhub install olmmlo-cmd/agent-guardrailsChạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.