Security & AI safety

AI guardrails & agent security

3,625 agents · most starred first

AI guardrails & agent security, page 5 of 37

agent-glovebox★ 52

You've probably been using Claude irresponsibly. Today, you can turn a new leaf.

pajaMAS★ 51

Multi-agent system (MAS) hijacking demos

crucible★ 51

pytest for AI agents - Autonomous red-teaming, behavioral monitoring & security testing for LLM agents

hermes-katana★ 51

Defense-in-depth security toolkit for LLM agents — taint tracking, proxy secret guard, policy engine, and red-team benchmarking

BackdoorAgent★ 51

BackdoorAgent is a stage-aware framework and benchmark that instruments LLM-agent workflows (planning, memory, tools) to systematically inject, track, and evaluate backdoor triggers across multi-step trajectories using unified metrics like ASR and clean accuracy.

AgentClaimGuard★ 51

Evidence gate for LLM agent claims - verify claims against evidence, tool results, and policies.

AEVS-sdk★ 50

Agent Execution Verification System — transparent audit SDK for AI agents.

ssh-agent-filter★ 50

filtering proxy for ssh-agent meant to be forwarded to untrusted servers

safety-for-agentic-ai★ 50

Improve safety, security, and privacy of AI systems at build, deploy and run stages.

key-connector★ 50

An agent that stores and provides cryptographic keys to Bitwarden clients.

Agentic-Guardian★ 50

[ICLR'26] Building a Foundational Guardrail for General Agentic Systems via Synthetic Data

g0★ 50

The control layer for AI agents. 1,200+ rules, 12 domains, 10 frameworks, 4,000+ adversarial payloads. Discover. Assess. Test. Monitor. Comply.

brin★ 50

credit score for context

SAFi★ 50

SAFi is an open-source runtime governance engine for agentic AI that enforces your policies, governs agent actions, and records every decision for audit.

cloakpipe★ 50

CloakPipe is the verifiable control plane for production AI agents: evaluate a release, certify it, enforce it at runtime, and prove what happened.

APIMitmHack★ 50

恶意中转,目标是支持对opencode,claudecode,openclaw等常见的agent的攻击。

prodsec-skills★ 50

Security skills for AI coding assistants and agentic systems

progent★ 50

Progent: Securing AI Agents with Privilege Control

keymaxxer★ 50

Secret manager for coding agents — run commands that use your API keys and tokens without ever exposing the secret to the model.

flar★ 50

FLAR is a Fast Light Agent Restrictor

cybersecurity-skills-zh★ 50

734+ 网络安全技能中文翻译版 | AI Agent 网络安全技能库 | agentskills.io 开放标准

siq-agent-security★ 50

Research artifacts for provenance-bound agent authorization, signed effect evidence, and reproducible security evaluation.

pam-ssh-agent★ 49

A PAM module that authenticates using the ssh-agent

arifos★ 49

arifOS — Constitutional MCP kernel for governed AI execution. AAA trinity: AGI (Mind) · ASI (Heart) · APEX (Judge). Built for the sovereign agentic era.

skill-audit★ 49

Audit agent skill definitions for security, completeness, and compatibility across Codex, Claude Code, OpenClaw, and more

HarnessAudit★ 49

Official codebase for the paper "Auditing Agent Harness Safety"

Agent-Security-Regression-Harness★ 49

Executable security regression testing for agentic applications and MCP-integrated systems.

ai-agent-hacking-writeups★ 49

A collection of public writeups on attacks against AI agents: prompt injection, tool abuse, memory poisoning, and data exfiltration.

ABeast★ 49

Lightweight exit monitoring and protection for AI agents.

oauth-agent-node-express★ 48

A Node.js OAuth Agent, providing API driven OAuth and OpenID Connect for SPAs

PromptGuard★ 48

Build production ready apps for GPT using Node.js & TypeScript

gap★ 48

Give AI agents secure access to your accounts - without sharing your credentials

GuanYu★ 48

JVM runtime class loading protection agent.(JVM类加载保护agent)

seccompagent★ 48

agent for handling seccomp descriptors for container runtimes

xfce-polkit★ 48

A simple PolicyKit authentication agent for XFCE

leash★ 48

Security guardrails for AI coding agents. Prevents accidental file operations outside working directory.

cmd-polkit★ 48

Easily create polkit authentication agents by using commands

clawpilot★ 48

🐾 ClawPilot — your OpenClaw copilot. Security-first agent skill for AI coding assistants. Supports 14+ chat platforms.

gpt-oss-safeguard-20b★ 48

gradio region:us

agentZ★ 48

Zero-trust agentic AI platform. Supports SaaS and OnPrem (airgapped) deployments.

ai-charter★ 48

Universal AI Charter: plain-language governance principles and a public certification registry for AI models, built on disclosure instead of a development slowdown.

awesome-llm-security★ 48

Agentic security, LLM security tools, research, and documents

agentguard★ 47

Universal security guardrails and workflow policies for AI coding agents.

llm-verify★ 47

Detect fake AI APIs — Verify if an LLM API is actually serving the model it claims. Catch resellers selling "Claude" or "ChatGPT" that secretly serve cheaper models. Behavioral fingerprinting with 32 forensic prompts.

psst★ 47

Beautiful (and themeable), unified UI for your pinentry, GNOME keyring-prompter and polkit-agent.

llm-firewall★ 47

AI Firewall & LLM security toolkit - protect your AI applications from prompt injection, jailbreaks, PII leakage, and adversarial attacks

agentmetry★ 47

Open-source security for AI coding agents: record, detect and block at the tool boundary, with a verifiable audit trail in the SIEM you already run.

agentdoor★ 47

The pre-auth layer for the agentic internet. Agents register, authenticate, and pay — programmatically. 3 lines of code. < 500ms

receipt-protocol★ 47

Minimal Ed25519-signed activity receipts for verifiable AI agents — self-attested, offline-verifiable, no blockchain.

agent.pw★ 47

Share APIs with agents, without sharing secrets.

agent-network-access-probe★ 47

Finds and fixes agent network access risks in OpenShell and Docker sandboxes.

cartai★ 46

The OS AI engineering and monitoring agent. 🦸‍♀️ Oversight and compliance copilot for trustworthy AI.

riphook★ 46

Deterministic security layer for Openclaw(Clawdbot), Cursor and Claude Code. Write secure code, prevent data exfil, and more

STARS★ 46

AI agent whose purpose is to conduct vulnerability tests on LLMs from SAP AI Core or from local deployments, or models from HuggingFace. The goal of this project is to identify and correct any potential security vulnerabilities.

agent-passport-system★ 46

Open AI agent governance protocol. Cryptographic identity, delegated authorization that narrows at every hop, gateway enforcement, and signed receipts. TypeScript, Apache-2.0.

pantheon-agent-polkit★ 46

Pantheon Polkit Agent

RepoGuardBench★ 46

Benchmarking repository-borne prompt injection attacks and lightweight defenses for local coding agents. DL4C @ ICML 2026.

HAIA★ 46

CC BY-NC-SA 4.0

pigeon★ 46

Delegated authority for AI agents. A Pigeon Pass says what an agent may do.

trace-spec★ 46

TRACE: Trust Runtime Attestation and Compliance Evidence. Open attestation standard for agentic AI governance.

cerbia★ 46

A modular threat detection library for AI and LLM pipelines. It provides language-agnostic guardrails through a configurable pipeline of loaders, multi-layer de-obfuscation, specialized scanners, and automatic enforcement.

quolo★ 45

Quolo enables companies to easily build their own platforms that leverage LLMs without the need to build custom permissions engines.

tacit★ 45

TACIT — Tracked Agent Capabilities In Types

opa-nginx-rbac★ 45

A proof of concept for RBAC on nginx with Open Policy Agent

dotnet-opa-wasm★ 45

Call Open Policy Agent (OPA) policies in WASM (Web Assembly) from .NET Core

envsitter-guard★ 45

OpenCode plugin that prevents agents/tools from reading or editing sensitive .env* files, while still allowing safe inspection via EnvSitter.

pytector★ 45

Easy to use LLM Prompt Injection Detection and Prompt Input Sanitization / Detector Python Package with support for local models, API-based safeguards, safeguard for tool calls, and LangChain guardrails.

clawmoat★ 45

The open-source agent firewall. Prevent AI agents from leaking data, using dangerous tools, and importing poisoned dependencies.

veritas_os★ 45

AI agent execution governance: pre-effect authorization, single-use permissions, policy enforcement, and verifiable reconciliation.

Armorer-Guard★ 45

Public SDKs and integration contracts for Armorer Guard.

Vulnerable-AI-Chatbot★ 45

An intentionally vulnerable AI chatbot to learn and practice AI Security.

jailbreak-fable★ 45

High-fidelity Claude Fable 5 (Mythos) environment emulation and automated multi-agent jailbreak (Pack Hunt) research laboratory.

pattern8★ 44

🎱 AI Agent Governance Framework — Constrain how AI Agents behave in your project. pip install pattern8

r19-iannuttall-claude-agents-security★ 44

🔒 Security & Compliance skill suite derived from iannuttall/claude-agents.

llm-confidentiality★ 44

Whispers in the Machine: Confidentiality in Agentic Systems

ai-protector★ 44

Ship AI agents with guardrails — not prayers. Self-hosted runtime protection for LLMs and tool-calling agents: block prompt injection, enforce tool permissions, redact sensitive data, and control what agents are allowed to do.

Makoto★ 44

When Claude says “tests pass”, makoto checks the record — every claim held against the agent's own logged deeds; fakes blocked, not warned. Each check ships at measured zero false positives. 誠

llm-guard-playground★ 44

docker region:us

agent-guard★ 44

Real-time secret-leak guardrails for AI coding agents (Claude Code, Codex), Git hooks, and CI.

agent-boundary-check★ 44

Verify what AI coding agents can actually read, write, execute and reach using synthetic canaries.

Perci★ 44

Fast, deterministic verifier and gate for AI agents: exact math, claim and citation checks, shell/write policy, hash-chained step ledger and a Temporal Router. Local MCP server + CLI in Rust. Preview.

RAGdrag★ 44

RAG pipeline security testing toolkit - 27 techniques across 6 kill chain phases, mapped to MITRE ATLAS

MasterKey★ 43

MASTERKEY is a framework designed to explore and exploit vulnerabilities in large language model chatbots by automating jailbreak attacks and evaluating their defenses.

Heimdall★ 43

Tamper-evident integrity monitor for the MCP config & server files your local AI agents load.

skill-sentinel★ 43

A security scanner for Agent Skill packages. Skill Sentinel uses multi-agent AI analysis to detect prompt injection, data exfiltration, command injection, malware, and other threats hiding in skill packages for Cursor, Claude Code, Codex, and OpenClaw.

Anthropic-API-Scanner★ 42

Scan GitHub for exposed Anthopic API Keys

RAGdrag★ 42

RAG pipeline security testing toolkit - 27 techniques across 6 kill chain phases, mapped to MITRE ATLAS

agent-lock★ 42

Confine an AI coding agent to one directory with a BPF-LSM program, and watch every escape it tries.

little-canary★ 42

Detects prompt injection by its effect on a sacrificial canary model, not just pattern matching: untrusted input hits a powerless model first, a behavioral check reads the residue, and it returns block, flag, or pass before your primary model acts. Inbound preflight sensor, not a guarantee.

ryk★ 42

Local guardrails for coding agents. Policy, approvals, 86 safety packs, OS sandbox, audit trail — no Docker.

council-of-ai★ 41

Security measure for agentic LLMs using a council of AIs moderted by a veto system. The council judges an agent's actions outputs based on specified categories.

MCPSecBench★ 41

MCPSecBench: A Systematic Security Benchmark and Playground for Testing Model Context Protocols

clawsecure-openclaw-security★ 41

Free OpenClaw security scanner. 3,000+ agents audited. 3-Layer Audit Protocol. OWASP ASI 10/10 coverage. AI agent integrity layer.

skillfortify★ 41

First formal security scanner for AI agent skills & plugins. Static analysis, supply chain verification, SBOM generation. 22 frameworks supported including MCP, LangChain, CrewAI.

awesome-claude-code-security★ 41

A awesome curated list of security resources, hardening tools, threat research, and governance frameworks specifically for Claude Code

Agent3Sigma-Canary★ 41

Agent3σ-Canary is an evaluation framework for AI Agent security in realistic runtime environments.

node★ 41

Runtime authorization for AI agents — Node.js SDK

antirea★ 41

Protect your project from AI-driven reverse engineering tools like REA. Agent instruction barriers, terminal shims, and build hardening in one skill.

UDora★ 40

[ICML 2025] UDora: A Unified Red Teaming Framework against LLM Agents

signet★ 40

Cryptographic trust and authorization evidence for AI agents. Every action a receipt, every grant a proof.

More in Security & AI safety