Tool
GitHub - rudratoshs/buried-injections: 🛡️ Regex catches 0%, Meta's Prompt Guard 2 catches 1% of 629 realistic AgentDojo injection attacks when they're buried in tool output. Reproducible benchmark. · GitHub
Publication date not yet evaluated · Added · 2 reports github.com
Page published
Coverage timeline
Why it matters
The 'buried-injections' benchmark exposes that widely used prompt-injection detectors, including Meta's Prompt Guard 2, fail almost completely against injections hidden in agent tool output, warning defenders not to rely on such filters for indirect prompt injection.
The 'buried-injections' GitHub project is a reproducible benchmark showing that regex-based detection catches 0% and Meta's Prompt Guard 2 catches roughly 1% of 629 realistic AgentDojo prompt-injection attacks when the malicious instructions are buried inside tool output (indirect prompt injection). The benchmark is published as runnable code demonstrating the gap between existing detectors and realistic indirect-injection scenarios.