{"repo":"raiph-ai/fireclaw","free":true,"listed":false,"github":"https://github.com/raiph-ai/fireclaw","clone":"git clone https://github.com/raiph-ai/fireclaw.git","description":"🛡️ A firewall for your agent's brain. Open-source security proxy that protects AI agents from prompt injection attacks. 4-stage pipeline, community threat feed, no bypass mode.","language":"JavaScript","stars":17,"topics":["ai-agents","ai-security","firewall","llm-security","oled","open-source","openclaw","prompt-injection","raspberry-pi","security-proxy"],"license":"AGPL-3.0","category":"ai-agents","readme_excerpt":"🛡️ FireClaw — A Firewall for Your Agent's Brain Open-source security proxy that protects AI agents from prompt injection attacks. Website • Quick Start • How It Works • Community Threat Feed • Want to Help? --- The Problem AI agents that browse the web are vulnerable to prompt injection attacks . Malicious websites can embed hidden instructions that hijack your agent's behavior — stealing data, executing commands, or overriding safety guidelines. Simple input filtering isn't enough; this is an adversarial problem that requires defense-in-depth. No existing open-source tool addresses this. FireClaw fills that gap. What FireClaw Does FireClaw sits between your AI agent and the internet. Every web fetch passes through a hardened 4-stage pipeline that strips prompt injection payloads before content reaches your agent's context window. Your agent calls FireClaw instead of fetching directly. FireClaw returns clean, factual content — no hidden instructions, no Unicode tricks, no encoding exploits. --- How It Works The Key Insight Even if the summarization LLM in Stage 3 gets injected, it has no tools, no memory, and no access to your data. It can only return text. And that text still passes through Stage 4 output scanning. The attacker is in a dead end. --- Features - 200+ Injection Patterns — Regex-based detection covering structural tricks, injection signatures, exfiltration attempts, and output manipulation - DNS-Level Blocklists — Integrates URLhaus, PhishTank, OpenPhish, and t","default_branch":null,"files":null,"tree":[],"storefront":"/r/raiph-ai","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/raiph-ai/fireclaw/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}