#red-teaming

Attack taxonomy, toolkit & defenses for persistent-memory poisoning of LLM agents. Author: Krishita Sanjay Choksi.

Dual-track AI reference architecture: RAG system design & AI Security Engineering (EN/ZH) | 双轨 AI 参考架构：RAG 系统设计与 AI 安全工程

EvalBot — local-first chatbot security & quality evaluation (FastAPI + Next.js). Evaluate chatbot answers against your own docs & guidelines with ML/NLP + AI-judge scoring. Apache-2.0.

A self-contained AI security lab demonstrating document poisoning, indirect prompt injection, and data exfiltration in RAG systems. Explores the "helpfulness paradox" across local and frontier LLMs.

Memory integrity for AI agents — Information-flow control against OWASP ASI06 memory & context poisoning.

OpenAI-compatible jailbreak proxy for Devin SWE-2 (server.codeium.com Connect-RPC) - system-layer override, category framing, invisible retry, RU/EN/ZH refusal handling

v0.8 preview · 2,288 multi-step agent trajectories · 513 hand-authored gold + 1,775 provenance-flagged augmented. A per-step benchmark that scores whether a verifier catches drift inside an agent's trajectory — not whether a prompt is harmful.

Curated research on AI agent social engineering, manipulated delegation, prompt injection, tool-use boundaries, and agent security evaluation.

AISRF, the AI Security & Research Framework: every LLM request from your apps and agents becomes a ticket a human approves before it reaches the backend. Analyzers, guardrails (Rebuff, LLM Guard, NeMo, Lakera), red teaming (garak, promptfoo, PyRIT), source code review, reports, MCP server.

How I validate LLM agents in regulated financial services: from regulation, the NIST AI RMF and its Generative AI Profile to research-based test design, with illustrations on simulated outcomes.

Does an LLM's willingness to blackmail depend on its persona — and on whether that persona is prompted or trained into the weights? Two-part study on Gemma-3-12B: 16 prompted personas (n=100) + 10 QLoRA persona adapters, in the Lynch et al. shutdown scenario.

🧪 Guimí Test AI — SDK Python para testes, observabilidade e conformidade de LLMs. LLM-as-Judge, Red-Teaming, LGPD/EU AI Act.

AI Agent Security: Multi-Step Tool Attacks — Kaggle (OpenAI) attack.py + working note