Agent Safety and Guardrails skills for Claude Code and Codex
Skills that keep agents inside their limits: governance and trust controls, guardrail frameworks such as NeMo Guardrails, LlamaGuard and Prompt Guard, prompt injection and supply-chain checks, MCP security audits, approval reviews before consequential actions, agent identity in Microsoft Entra, and Claude Code hooks that block destructive commands. They give an agent named rules for what it may do.
Highest-scoring agent safety skills in the October 22, 2026 snapshot: skill-scanner from getsentry/skills (88/100), hosted-agents from muratcankoylan/Agent-Skills-for-Context-Engineering (86/100) and aig-scanner from Tencent/AI-Infra-Guard (82/100). Scores come from the Suede Skill Check.
- 27 listed skills
- 14 repositories
- 4 with full pages
16 from official publishers. 5 bundle scripts. 3 connect to an outside service. Listed in AI Agents and LLM Engineering (21), Security (4), Developer Tools and Frameworks (2).
All 27 agent safety skills
No matching skills. Try fewer words or clear your search.
| Skill | Repository | What it does |
|---|---|---|
| aig-scanner 82/100 · B | Tencent/AI-Infra-Guard | Runs AI security scans of infrastructure, AI tools and skills, and AI agents, and evaluates LLM jailbreak resistance with Tencent's AI-Infra-Guard. |
| hosted-agents 86/100 · B | muratcankoylan/Agent-Skills-for-Context-Engineering | Guides design of hosted or background agent infrastructure, including sandboxed execution, remote coding environments, warm pools, session persistence, multiplayer collaboration, and self-spawning agents. |
| safe-browser 75/100 · C | browserbase/skills | Builds constrained-browser agents whose safe_browser tool enforces a domain allowlist, so an agent can browse without raw browser, shell, or CDP access. |
| skill-scanner 88/100 · B | getsentry/skills | Scans agent skills for prompt injection, malicious scripts, excessive permissions, secret exposure, and supply chain risks before installation. |
| agent-governance 67/100 · D | github/awesome-copilot | Patterns and techniques for adding governance, safety, and trust controls to AI agent systems. Use this skill when: - Building AI agents that call… |
| agent-supply-chain 80/100 · B | github/awesome-copilot | Verify supply chain integrity for AI agent plugins, tools, and dependencies. Use this skill when: - Generating SHA-256 integrity manifests for agent… |
| agents-harden 74/100 · C | aws/agent-toolkit-for-aws | Use when preparing your agent for production — IAM scoping, inbound auth (JWT, SigV4), secrets management, cold start optimization, session… |
| ai-prompt-engineering-safety-review 69/100 · D | github/awesome-copilot | Comprehensive AI prompt engineering safety review and improvement prompt. Analyzes prompts for safety, bias, security vulnerabilities, and… |
| ax-audit 87/100 · B | mblode/agent-skills | Audits agentic products for tool parity, authority, approval payloads, recovery, and trust using 24 rules and a ship verdict. Use when asked for an… |
| careful 70/100 · C | garrytan/gstack | Safety guardrails for destructive commands. (gstack) |
| codex-host 69/100 · D | avelikiy/great_cto | Run the great_cto controlled Codex lifecycle with controller-owned writes, verifier evidence, human gates and optional artifact release. |
| constitutional-ai 74/100 · C | Orchestra-Research/AI-Research-SKILLs | Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then… |
| deep-agents-orchestration 69/100 · D | langchain-ai/langchain-skills | INVOKE THIS SKILL when using subagents, task planning, or human approval in Deep Agents. Covers SubAgentMiddleware, TodoList for planning, and HITL… |
| entra-agent-id 84/100 · B | microsoft/azure-skills | Provision Microsoft Entra Agent Identity Blueprints, BlueprintPrincipals, and per-instance Agent Identities via Microsoft Graph, and configure OAuth… |
| entra-agent-user 69/100 · D | github/awesome-copilot | Create Agent Users in Microsoft Entra ID from Agent Identities, enabling AI agents to act as digital workers with user identity capabilities in… |
| ethos-explore 79/100 · C | NVIDIA/skills | Captures what an AI agent is supposed to do — its purpose, boundaries, and success criteria — as a durable snapshot for anyone building, evaluating… |
| ethos-orchestrate 85/100 · B | NVIDIA/skills | Entry point for an AI agent's ethos.md: a durable snapshot of what the agent is for, what it must not do, and what success looks like. Routes a… |
| ethos-write 79/100 · C | NVIDIA/skills | Writes an AI agent's ethos.md — a durable snapshot of its purpose, intended behavior, boundaries, and success criteria — from the answers gathered… |
| git-guardrails-claude-code 77/100 · C | mattpocock/skills | Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to… |
| harness-engineering 81/100 · B | muratcankoylan/Agent-Skills-for-Context-Engineering | This skill should be used when designing autonomous agent harnesses: research loops, evaluation scaffolds, locked and editable surfaces, durable… |
| llamaguard 66/100 · D | Orchestra-Research/AI-Research-SKILLs | Meta's 7-8B specialized moderation model for LLM input/output filtering. 6 safety categories - violence/hate, sexual content, weapons, substances… |
| mcp-implementation-security-review 76/100 · C | github/awesome-copilot | Review the implementation source code of MCP (Model Context Protocol) servers, clients, and tool handlers against a security baseline —… |
| mcp-security-audit 78/100 · C | github/awesome-copilot | Audit MCP (Model Context Protocol) server configurations for security issues. Use this skill when: - Reviewing .mcp.json files for security risks… |
| nemo-guardrails 66/100 · D | Orchestra-Research/AI-Research-SKILLs | NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input/output validation, fact-checking, hallucination… |
| nemotron-policy-generator 75/100 · C | NVIDIA/skills | Generates BYO custom safety policies for NVIDIA Nemotron content-safety guardrails — Nemotron-Content-Safety-Reasoning-4B (text) and multimodal… |
| prompt-guard 66/100 · D | Orchestra-Research/AI-Research-SKILLs | Meta's 86M prompt injection and jailbreak detector. Filters malicious prompts and third-party data for LLM apps. 99%+ TPR, <1% FPR. Fast (<2ms GPU)… |
| verify-agent-action 77/100 · C | github/awesome-copilot | Review a proposed AI-agent action or human-approval packet before execution. Use when an agent wants to run a consequential tool, command… |
How to choose agent safety skills
Separate runtime guardrails from review skills. NeMo Guardrails, LlamaGuard and Prompt Guard filter inputs and outputs in a running app; agent-governance, ax-audit and mcp-security-audit review a design or configuration once.
For coding agents, the lightest protection is a hook. git-guardrails-claude-code and careful block dangerous commands before they run, with no service behind them.
Identity skills (entra-agent-id, entra-agent-user) change tenant settings. Run them with an admin who owns that tenant and keep the resulting permissions as narrow as the task needs.
Ranked agent safety shortlists
Related topics
Security Auditing and Review 108AI Agent Builders 56Azure 78Claude Code Setup and Hooks 31Dependency Management 23MCP Servers 82Prompt Engineering 28Vulnerability Scanning 55
Find your next useful skill
Explore Rising Skills with quality scores and source reports, or search from inside your agent.
Get Rising Skills by email · Read Rising Skills · Install the finder · State of Agent Skills