# team blackbox

- **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:** [Nabid Nur](https://cerebralvalley.ai/u/navi12)
- **GitHub:** https://github.com/Hisernberg/spanguard/
- **Demo video:** https://youtu.be/ur34rPdMIME
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3175

Your AI agent just read a shipping notification. Buried in the carrier note is an
instruction to email your customer list to an attacker. The agent cannot tell data from
instruction — and whole-document guards can't either: they block everything (agent blind)
or pass everything (agent hijacked).

SpanGuard is the third option: strip the attack, keep the data. An OpenAI-compatible
security gateway: tool results, pages, MCP payloads and RAG chunks pass through before the
agent reads them. Validated rules + a 135M multilingual span model + a family of
sub-million char models find the injection spans and replace them with [stripped]
markers — order IDs, totals, names and JSON structure survive untouched.

v6 adds three measured layers: a split-conformal bound (over-stripping bounded by
construction — 0.8% on 500 fresh adversarial docs), a task-conditioned gate (canonical
AgentDojo attack escape 80.6% → 40.3% with the user's task given), and an agent-loop
harness that scores outcomes, not vibes: utility 100% with hijack 1.3% — the best
combination of four systems. A semantic judge (an LLM reading the cleaned output) audited
the rules: 30/30 verbatim-stripped, 0/30 semantic escapes.

Everything is open: 7 languages, every weight on Hugging Face, every run record committed
and re-verifiable, honest costs published next to every win. CPU-only, ~21 ms per document,
no LLM call in the guard path.

## More from Meta Global AI Developer Hackathon

- [TITANS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3172)
- [KP419](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3173)
- [Keyboard Repair Check](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3174)
- [Hai Family](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3176)
- [SheetPilot](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3177)
- [icarus](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3178)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3175. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt