Mobile GUI Agents under Real-world Threats: Are We There Yet?

This project, emerging from a very recent research paper (3 days old), addresses a critical bottleneck for the adoption of mobile AI agents: security. While it has 0 stars, the 8 forks indicate immediate interest from the academic community. The core value is the evaluation methodology for 'Agentic Security' in a mobile context. However, the defensibility is low because this is primarily a research artifact rather than a product or a hard-to-replicate infrastructure. Frontier labs like Apple (with Apple Intelligence) and Google (with Gemini/Android integration) are actively building their own internal red-teaming and safety layers for GUI agents. As these platforms control both the OS and the model, they are likely to absorb the security patterns described here into their proprietary safety filters. This project serves as a vital 'warning shot' for the industry but lacks a moat beyond its initial dataset and findings, which will likely be superseded by more comprehensive industry-led safety benchmarks within 6 months.