ASL V6

Open-Source AI Security Red-Teaming Engine

10 followers

Open-Source AI Security Red-Teaming Engine

10 followers

Standard SAST tools flood developers with false positives on LLM code. ASL V6 is an open-source vulnerability scanner for AI agents that uses Python AST analysis to ignore test noise, and ephemeral Docker containers to verify if prompt injections are actually exploitable in runtime. It reduces false positives by 98%. Built by researchers with 6 CVEs in production AI frameworks. 100% free and local.

Free

Launch tags:Open Source•GitHub•Development

Launch Team

CloudflareStill paying to deploy your apps? Try Cloudflare Workers for free

Promoted

Maker

📌

Hey Product Hunt! 👋 I’m Siva Aditya, a security researcher with 6 CVEs in production AI frameworks like AutoGPT and Flowise. I built ASL V6 because I was tired of standard security tools throwing thousands of false alerts when scanning LLM orchestration code. ASL V6 uses abstract syntax tree (AST) filtering to ignore test suites and docstrings, then tests untrusted payloads inside an isolated Docker container to see if they actually execute. I ran it across LangGraph, Rasa, and Weaviate this week, and it reduced raw pattern noise by over 98%. The engine is 100% free and runs locally. I open-sourced it because AI security shouldn't be hidden behind massive enterprise paywalls. (P.S. If you are building AI agents in production and don't want to run the infrastructure yourself, my lab offers a one-time $1,500 architecture assessment. Details are on the GitHub README!) I'll be hanging out in the comments all day to answer any technical questions about AST analysis or agent exploits!

Report

3mo ago

One thing that would help a lot is adding a GitHub Action that auto-runs ASL V6 on PRs and posts the real findings as inline comments, since most teams won't remember to run it locally before pushing. Would make adoption way smoother.

Report

3mo ago

Maker

@ira_jett this is exactly why I launched here — got this feedback 3 days ago and shipped it last night.

v6.1.0 has the GitHub Action now. Inline comments + merge blocking on high severity:

- name: ASL V6 Auto-Auditor
  uses: sivaadityacoder/asl-v6@v6.1.0
  with:
    github_token: ${{ secrets.GITHUB_TOKEN }}
    fail_on_high_severity: 'true'

Would be super grateful if you kick the tires on it and tell me what's broken. Real feedback > everything.

Report

3mo ago

Reviews