Go to Apify Store
User picture

Shokhabbos Mukhammatov

shohabbos_m

Security testing for AI agents that handle money. I build Fire Drill: drills an accounts-payable agent and shows where a control failed.

ACTOR STATS

2 public Actors

3 total users

1 monthly user

>99% runs succeeded

Hello 👋

I build tools that test whether AI agents hold up under pressure.

Fire Drill is what I'm working on now: a controls-testing harness for accounts-payable AI agents. An AP agent can read invoices, look up vendors and move money, so the interesting question isn't whether it works — it's what it does when someone asks it to change a vendor's bank details. Fire Drill runs that as a drill, shows the turn where a control failed, and scores the result from ledger state rather than from what the agent says it did.

Two Actors, meant to be used together:

  • 🔥 fire-drill  — runs the drills and writes the audit report.

  • 🎯 ap-practice-target  — a sandboxed AP agent you can drill over HTTP, with a mock ledger and fictional vendors.

  • 🎓 San Francisco State University

  • 🏗️ Built solo at Cal Hacks 13

  • 📬 smukhammatov@sfsu.edu

Public Actors