magicmadeint

magicmadeint / wipedown

Public

Zero-Trust Semantic Scraper & Firewall Proxy for AI Coding Agents. Intercepts and neutralizes web-scraped prompt injections locally before they hit your terminal window.

magicmadeintelligence.com agentic-workflow ai-security aider cursor-agent developer-tools
12
1
85% credibility
Found Jun 01, 2026 at 12 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

WipeDown is a security tool that protects AI coding assistants from prompt injection attacks hidden in webpages. It acts like a filter that checks websites for malicious instructions before letting your AI assistant read them. The tool runs locally on your computer, automatically discovers your existing AI setup, and can be used either as a background protection layer that your assistant routes through, or as a manual tool you use yourself to check specific pages or files. It combines fast pattern detection with AI-powered analysis to neutralize threats while preserving useful content.

How It Works

1
🔍 You hear about a security tool for AI assistants

A friend tells you about WipeDown, a tool that protects your AI coding assistant from sneaky hidden commands on websites.

2
🛡️ You install WipeDown on your computer

You download and set up WipeDown in just a few simple steps, and it automatically finds the AI model you're already running locally.

3
🌐 You point your AI assistant through WipeDown

Instead of letting your AI assistant visit websites directly, you route it through WipeDown's local protection layer.

4
⚡ WipeDown catches hidden attacks before they reach your assistant

When a webpage tries to sneak in malicious instructions, WipeDown spots them and neutralizes them automatically while keeping the useful content safe.

5
You can also use WipeDown manually whenever you want
💻
You check a webpage directly

You ask WipeDown to fetch and clean a specific URL, and it returns safe, readable content.

📄
You scan a local file

You can also scan files stored on your computer to make sure they're safe before sharing.

6
📝 Your assistant gets clean, safe information

The cleaned content is saved as a neat document that your AI assistant can safely read and learn from.

🎉 Your AI assistant stays safe and helpful

With WipeDown watching the door, your AI coding assistant can explore the web confidently without being tricked by hidden attacks.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 12 to 12 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 wipedown?

WipeDown is a security-focused web scraper built in Python that acts as a local firewall between untrusted web content and your AI coding agent. It fetches pages, strips malicious formatting and hidden injection payloads, then passes the sanitized content through a local LLM to neutralize imperative commands before they reach tools like Cursor, Claude Code, or Cline.

The tool runs as a local HTTP proxy on port 8010, so you point your agent's fetcher at it and it handles the rest. It also ships a full CLI with commands for one-off fetches, automated configuration, and system validation.

Why is it gaining traction?

Prompt injection via scraped web content is a real and growing threat as more developers wire autonomous agents directly to the web. WipeDown addresses this with a two-stage defense: a fast signature-based scanner that catches obvious jailbreak patterns before any LLM call, followed by semantic sanitization that rewrites hidden commands into safe passive text.

The hook is the agent auto-configuration mode. Running `wipe down configure --auto` scans your local machine for running inference servers like llama-server or Ollama and writes a ready-to-use configuration file in seconds. No manual setup, no guesswork about endpoints. It also handles Twitter/X links by routing through open-source proxy mirrors, which developers fetching technical threads desperately need.

Who should use this?

If you run autonomous coding agents that browse the web or process scraped content, this is directly relevant to your threat model. Security-conscious developers evaluating zero trust assessment tool options on GitHub for their AI pipelines will find this fills a specific gap that generic web scrapers ignore. Researchers building workflows that ingest tweets or articles with agent chains should test it in staging before production use.

Verdict

At 12 stars, WipeDown is early and experimental, with a credibility score of 0.85% and no visible test suite in the repository. The documentation is solid and the concept is sound, but the project needs real-world battle testing before trusting it with production agent pipelines. Try it locally, contribute test cases, and watch it mature. Do not enable auto-approve modes on agents consuming sanitized output regardless.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.