Agent Audit
A security audit service for companies' AI chatbots and agents: it runs real attacks and delivers a report with evidence and concrete fixes.
agent-audit-ten.vercel.app
WHAT IT DOES
Most teams ship an AI chatbot without ever testing it against prompt injection. If the assistant reads e-mails, documents or web pages and can call tools, anyone who writes to it can try to steer it.
Agent Audit tests the system for prompt injection, jailbreaks, data exfiltration, tool misuse and system-prompt leakage. Each response is scored for evidence that an attack worked, and the result is a PDF report with an executive summary, a severity table, evidence and fixes.
Testing only starts once the system owner has given written permission.
$0
the mini-scan is always free. Full audit $400–800, monthly monitoring $150–300.
HIGHLIGHTS
- Free mini-scan: 10–15 attack probes and a one-page summary.
- Full audit ($400–800): the full attack library, multi-turn scenarios, a PDF report and a 30-minute call.
- Monthly monitoring ($150–300): catches regressions when prompts or models change.