GoPlusSecurity

GoPlusSecurity / agentguard

Public

Security guard for AI agents — blocks malicious skills, prevents data leaks, protects secrets. 24 detection rules, runtime action evaluation, trust registry.

293
46
100% credibility
Found Feb 05, 2026 at 36 stars 8x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

AgentGuard protects users of AI coding agents by scanning skills for security risks and blocking dangerous runtime actions like destructive commands or data leaks.

How It Works

1
😟 Worry about AI risks

Your helpful AI coding assistant can run commands on your computer, but bad instructions might make it delete files or steal private info.

2
🔍 Find AgentGuard

Discover this friendly security guard made just for AI users to keep everything safe from sneaky tricks.

3
📥 Easy setup

Download the tool and run its one-click installer – it adds protection to your AI setup in moments.

4
🛡️ Protection turns on

Watch as it automatically checks new AI helpers for dangers and blocks harmful actions before they happen – feel instantly safer!

5
🔍 Check any helper

Tell your AI to scan a folder of code, and get a clear report on risks like hidden tricks or bad patterns.

6
Pick your safety level
🔒
Strict mode

Blocks anything even slightly risky – perfect for high caution.

⚖️
Balanced mode

Stops dangers outright and asks before risky steps – great for daily use.

🔓
Permissive mode

Only stops the worst threats – for when you want smooth flow.

😌 Code safely forever

Now relax knowing your AI can't harm your computer or steal secrets – happy, protected creating!

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 36 to 293 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is agentguard?

AgentGuard is a TypeScript security guard for AI agents on platforms like Claude Code, Cursor, and GitHub Copilot. It blocks malicious skills from stealing keys or draining wallets, prevents data leaks via 24 detection rules, and adds runtime verification of AI agents through action evaluation and a trust registry. Users get automatic skill scanning on load, protection levels (strict, balanced, permissive), and commands like `/agentguard scan ./path` or `/agentguard action "curl | bash"`.

Why is it gaining traction?

Real-time hooks catch destructive commands like `rm -rf /` or webhook exfil before execution, plus Web3 checks via GoPlus API flag unlimited approvals and malicious contracts. Unlike basic linters, it tracks skill-initiated actions and enforces per-skill capabilities, handling github security alerts in agent sessions. Agent Skills compatibility means one npm install protects across CLIs without workflow changes.

Who should use this?

Claude Code users installing random skills from GitHub repos. Cursor or security GitHub Copilot devs needing agent guard against prompt injection. Web3 agent builders verifying tx/sign actions runtime to avoid wallet risks.

Verdict

Try it if you're deep in AI agents—setup.sh makes Claude integration dead simple, docs cover MCP servers and SDK use. At 80 stars and 1.0% credibility, it's early-stage; audit hooks yourself for production. Strong start for github security scanning in agent workflows.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.