# MixShield

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Huzaifa Pathan](https://cerebralvalley.ai/u/HP007)
- **GitHub:** https://github.com/HP0077/mixshield
- **Demo video:** https://www.youtube.com/watch?v=uiouI_9ESK0
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1060

AI agents read emails, chats and documents, and attackers can hide instructions in that text (indirect prompt injection). The attacker chooses the language. MixShield tests and protects tool-using agents against hidden instructions written in English, Hinglish (Hindi and English mixed, typed in English letters) and Hindi.

It has two parts:
1. A benchmark on AgentDojo: 300 attacks and 450 benign texts in three languages, plus 30 adaptive attacks. On the test split, Llama Guard 3 caught 84.3% of attacks in English but only 68.6% in Hinglish (McNemar p = 0.000145), and Llama Prompt Guard 2 at its default threshold caught 0 of 70 in every language.
2. An open-source guard (Python library + MCP proxy) with two checks: a screen that checks every tool output (Llama Prompt Guard 2 + Muse Spark Open as a code-mixing-aware classifier that gives a reason), and an action gate that checks every payment or message before it runs (allow / ask / block).

The MixShield screen caught 99.5% of Hinglish attacks with 0.0% false alarms. With a worst-case simulated agent, MixShield stopped 189 of 189 attacks, while Muse Spark kept 88.5% of real tasks working (92.3% with no defence). Limits are listed openly in the README (simulation, 2 of 27 adaptive attacks got through in default mode).

## More from Meta Global AI Developer Hackathon

- [ShieldPocket](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1057)
- [ReflectAI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1058)
- [Jae Hun Hwang](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1059)
- [ZeLabs Ltd](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1061)
- [DrakenMario](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1062)
- [Nichita Stane](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1063)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1060. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt