barbarkaragul-oss

Agentwall — AI skill for Claude Code

AI community

Zero-Dependency Runtime Firewall, Secret DLP & Process Sandbox for AI Agents and MCP Servers.

How to install Agentwall

This entry records only its repository, not the path inside it, so there is no exact command to give. Open barbarkaragul-oss/agentwall and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agentwall does

Zero-Dependency Runtime Firewall, Secret DLP & Process Sandbox for AI Agents and MCP Servers.

Alternatives in AI

WindsurfAPI — Turn Windsurf / Devin Desktop's 100+ AI models (Claude, GPT, Gemini, DeepSeek, Kimi, GLM, SWE) into OpenAI-, A 3k ★
Hol Guard — Open-source antivirus for AI agents: block risky tools, secret access, prompt injection, malicious packages, M 485 ★
Prismor — Self-hosted runtime control plane for AI agents 334 ★

README

:shield: AgentWall

**Zero-Dependency Runtime Firewall, Secret DLP & Process Sandbox for AI Agents and MCP Servers**

[![CI Status](https://github.com/barbarkaragul-oss/agentwall/actions/workflows/ci.yml/badge.svg)](https://github.com/barbarkaragul-oss/agentwall/actions) [![Release](https://img.shields.io/github/v/release/barbarkaragul-oss/agentwall?color=brightgreen)](https://github.com/barbarkaragul-oss/agentwall/releases) [![Go Report Card](https://img.shields.io/badge/go%20report-A+-brightgreen.svg)](https://goreportcard.com/report/github.com/barbarkaragul-oss/agentwall) [![Go Reference](https://pkg.go.dev/badge/github.com/barbarkaragul-oss/agentwall.svg)](https://pkg.go.dev/github.com/barbarkaragul-oss/agentwall) [![Zero Dependencies](https://img.shields.io/badge/dependencies-0-blue.svg)](#) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)

*Stop autonomous AI agents (Claude Code, Cursor, OpenClaw, MCP Servers) from leaking your `.env` secrets, reading `~/.ssh/id_rsa`, executing `rm -rf /`, or falling for Indirect Prompt Injections.*

:zap: Why AgentWall?

Developers are giving autonomous AI agents and third-party **Model Context Protocol (MCP)** servers full access to their local machines. A single hallucination or malicious prompt hidden inside a webpage or repository can cause an agent to:

Exfiltrate Secrets: Leak AWS_ACCESS_KEY_ID, GitHub PATs (ghp_), Slack tokens, or OpenAI (sk-) keys into external LLM context windows.
Access Sensitive Files: Silently read ~/.ssh/id_rsa, ~/.aws/credentials, ~/.kube/config, or /etc/shadow.
Execute Destructive Commands: Run rm -rf /, reverse shells, or unauthorized network calls.

**AgentWall** wraps any CLI agent or MCP server with a **single zero-dependency Go binary** and inspects `stdio` (JSON-RPC 2.0) traffic and OS-level pr