Show HN: OpenAPPA – open-source deterministic guardrails that don't break agents

OpenAPPA, an open-source deterministic guardrail, claims zero successful prompt-injection data leaks while keeping agent task completion near 90 percent.

OpenAPPA is an MIT-licensed guardrail that runs outside an agent loop and assigns each data flow an audience-and-trust label, then derives allow or deny decisions algebraically. The project says this stops prompt-injection data exfiltration that a second classifier can miss, while sanitizers, one-time approvals, and isolated subagents keep agents usable. Its benchmark claims 89% task completion and 0% successful attacks, compared with 90% and 10% for Claude Auto mode and 41% and 31% for Microsoft FIDES. The post argues probabilistic judges cannot track data across tool calls and still leave residual failures at scale.