SMALL MIND

Work

Everything, with receipts.

Red-team results scored by someone else's judges, research frozen before it was analyzed, and the code behind all of it. If a claim here can't be checked, it isn't here.

Gray Swan IPI Black, Wave 1
#14 of ~1,600 · top 1%
JailbreakBench, judge swapped
Kendall's tau 0.115
Published, with receipts
4 papers · 4 public repos

Red team

Gray Swan Arena, IPI Black

Indirect prompt injection against safeguarded AI agents working in simulated Gmail, Slack and GitHub. Wave 1 finish: #14 of roughly 1,600 participants, 10 accepted breaks, every one graded by the arena's own judges rather than by me.

IPI Field Manual

How indirect prompt injection works against agents, the benchmarks that measure it, the incidents on record, and where defenses break. Built from a 19-report, 150+ source research corpus.

Small Mind Security assessments

Fixed-scope red-team assessments for LLM and agent products: prompt injection, agent and tool abuse, data exfiltration and RAG poisoning, delivered as a findings report.

Research

Code