Guardrails for Claude Code
github.com/ahmed-alstaty/guardrails-plugin- Category
- AI Agents
- Rank
- No. 2240Tools index
Previous survey · No. 2124 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Security & Identity
- Interfaces
- Agent Skill / Plugin
- GitHub
- 2 stars
- Latest release
- v1.0.1
- Date
About
Guardrails is a free Claude Code plugin that denies destructive shell commands and edits to secrets or protected paths before the tool runs, and requires approval for deployment and infrastructure files. At session start it prints an OK / WARN / SKIPPED report of loaded instruction files, rules, skills, hooks and plugins, and it lints CLAUDE.md files. It uses plain python3 scripts with no dependencies.
At a glance
Guardrails for Claude Code adds hooks that deny or ask before selected shell commands and file edits. It is worth trying if you want a repository policy for recurring actions such as force-pushing a shared branch or changing deployment files.
Good fit
Claude Code users who want inspectable rules and a report of the instructions and hooks loaded at startup.
Consider something else when
Skip it if your goal is to isolate the process or prevent an agent from reading secret files.
Start here
In a scratch project, run the plugin’s guardrails-test dry run on one harmless command and one command matching a protected rule. Check the allow, ask or deny result without executing either command.
Cost and requirements
MIT-licensed. Requires Claude Code with plugin support and Python 3.8 or later on PATH. The hooks use no additional Python packages.
Limits to know
The hook judges supported tool calls using command and path rules. It leaves reads unblocked and lets configured exceptions change the outcome. Check the coverage before relying on it for a project.
Related work
- Anthropic, maker of Claude Code (context)
Sources checked 2026-11-05. Source-based assessment.
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.