claude

guardrails

Data-driven guardrails for agent Bash commands: deny or warn rules with per-command retry, session modes an agent may enable only when you say so, global and per-project state, and presets (docs-first, process-safety, modern-cli, shell-hygiene) installed by a setup skill.

0 on GitHubv0.26.0updated 24 days agodevelopment
/

Plugin guardrails from marketplace gaetans-claude-plugins in gaetschwartz/claude-plugins. Commands read from the repo's manifest — 23 days ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (8)

  • audit

    Use only when the user explicitly asks to audit or review whether guardrails rules denied commands correctly, e.g. "audit the guardrails denials", "were the no-pkill denials justified?", "did any guardrails rule block something it should not have?". Reads recent denials from the Claude Code transcri

  • edit

    Use when the user explicitly asks to change, disable, enable, reword or remove an existing guardrails rule, e.g. "make the strings rule a warning", "turn off the pkill rule in this repo", "remove no-pkill". Never use it to get past a guardrails denial.

  • explain

    Use when someone asks why a command was denied or warned on, why a command was not caught, or what a guardrails rule covers, e.g. "why was my command denied?", "what does the no-pkill rule cover?", "why doesn't this rule catch curl | sh?". Callers, including other agents, must pass the rule id or th

  • mode

    Use when the user says this session is a specific kind of work covered by a guardrails mode, e.g. "we're reverse-engineering this binary", "this is RE work, strings is fine here", "turn on reverse-engineering mode", or asks to switch such a mode on or off, or to declare or undeclare one. Never enabl

  • new

    Use when the user explicitly asks to create a new guardrails rule, e.g. "block pkill for agents", "write a rule that warns on curl | sh", "add a guardrail for X". Interviews the user, tests the rule on examples and edge cases, shows it for confirmation, then writes it. Never use it to get past a gua

  • setup

    Use when the user wants to set up guardrails, install recommended guardrails presets, or asks which guardrails suit the work done on this machine or in this repo, e.g. "set up guardrails", "install the process-safety preset", "I do reverse engineering in this repo, configure guardrails for that". Ne

  • stats

    Use when the user asks how often guardrails rules fired or how long they take, e.g. "how often did each guardrails rule fire?", "which rules are slow or never fire?", "show guardrails stats for no-pkill". Counts and timings only, no explanation; use guardrails:explain for why a rule did or did not c

  • status

    Use when the user asks what guardrails rules, modes or problems are active, e.g. "what guardrails are active?", "list my guardrails rules", "is the managed file in use?", "are any guardrails rules suspended?". A plain list, no explanation; use guardrails:explain for why a rule did or did not catch a

Hooks (2)

  • PreToolUse

    sh "${CLAUDE_PLUGIN_ROOT}/hooks/guardrails.sh"

  • SessionStart

    sh "${CLAUDE_PLUGIN_ROOT}/hooks/guardrails.sh" session-start

View source on GitHub

More from gaetschwartz/claude-plugins

All 9 plugins →

chezmoi

claude

chezmoi dotfile-manager expertise — source-state attribute grammar, templating, scripts, externals, encryption and secrets — plus update/push skills that resolve sync conflicts autonomously and a hook that flags source/destination drift.

9 skills · hooks

0

dioxus

claude

Dioxus 0.7 expert: local docs, example and source search, Serena MCP for Rust symbols, and a dioxus-expert subagent that cites its sources.

1 skill · 1 agent · 1 MCP server

0

gdt

claude

Gaëtan's dev tooling — personal development-workflow skills. sdd: autonomous spec-driven development (worktree → specs → plan → adversarial review → execution). memory-optimization: fan-out audit that fact-checks a Claude Code memory store against the live system, consolidates it, and rebuilds MEMORY.md. worktree-sweep: fans a diagnostics backlog (clippy lints, compiler warnings) out to one worktree agent per rule family, lands the fixes linearly, verifies, and cleans up.

3 skills

0

p4

claude

P4 / Tofino SDE SME: bundled source-grounded P4 corpus (P4_16 spec, TNA, BF-Runtime, SDE guides, tutorials) served via a knowledge-rag MCP, and a p4-expert subagent that cites every answer for Q&A, codegen, and review across the full P4 stack (p4c/bf-p4c, P4Runtime/BF-RT, open-p4studio, bmv2/Tofino).

1 skill · 1 agent · 1 MCP server

0

tgrep

claude

Search with tgrep (Microsoft's trigram-indexed, ripgrep-compatible grep) instead of the Grep tool — a SessionStart hook keeps a tgrep server warm per repository without double-serving, a search skill that loads live index state and gets command shape, scoping and freshness right, and a hook skill to disable, exclude a repo, or stop a server.

2 skills · hooks

0

Learn a new language with a Claude-powered tutor. The tutor can help you practice reading, writing, and speaking in your target language, and provide feedback on your progress.

1 skill

0

Similar plugins

superpowers

claudecodex

Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques

15 skills · hooks

obra/superpowers 316,836

CMEM Pro

claude

CMEM Pro for the Claude app - hooks stream tool use to cmem.ai and inject observations into new sessions and agents

3 skills · 1 MCP server · hooks

thedotmack/claude-mem 116,618

context7

claudecodex

Up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your coding agent's context.

1 MCP server

upstash/context7 64,709

mempalace

claudecodex

AI memory system — mine projects and conversations into a searchable palace. 47 MCP tools, auto-save hooks, guided setup.

3 skills · 6 commands · 1 MCP server · hooks

MemPalace/mempalace 60,536

Build powerful, stateful CLI interfaces for any GUI application using the cli-anything harness methodology.

5 commands

HKUDS/CLI-Anything 58,665

Automates browser interactions for web testing, form filling, screenshots, and data extraction

12 skills

vercel-labs/agent-browser 50,373

Maintain guardrails?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins9 plugins