AIUC-1 requires quarterly third-party adversarial testing. We built the open-source framework that does it. 209 tests. 4 protocols. 20 enterprise platforms.
UiPath certified March 9. Cisco is a technical contributor. Phil Venables (former Google Cloud CISO) calls it "a SOC 2 for AI agents." Your procurement team is going to ask about this.
Adversarial robustness (B001), detect adversarial input (B002), input filtering (B005), output over-exposure (B009)
Restrict unsafe tool calls (D003), third-party testing of tool calls (D004) across MCP, A2A, L402, x402
Risk taxonomy (C001), pre-deployment testing (C002), third-party testing for harmful/out-of-scope outputs (C010, C011)
Assign accountability (E004), vendor due diligence (E006), audit logging (E015)
Limit agent data collection (A003), protect IP & trade secrets (A004)
Prevent AI cyber misuse via GTG-1002 APT simulation (F001)
The framework is open source, pip installable, zero dependencies. Test your own agents before the auditor does.
Then: agent-security test mcp --url http://your-server/mcp --trials 10
We run the 209-test harness against your AI agent deployment, map the results to AIUC-1 requirements, and deliver a gap analysis with remediation priorities. You get audit-ready evidence for B001, C010, and D004.
Book a 30-Minute Assessment CallOr email: mspro3210@gmail.com