
Vigil
AI harness for long-running code audits, web pentests, and bug bounty hunting, built around how real investigations fail: anchored hypotheses, false…

AI harness for long-running code audits, web pentests, and bug bounty hunting, built around how real investigations fail: anchored hypotheses, false…

Vulnerable app with examples showing how to not use secrets

Security research on context compaction in coding agents: shows how forged role labels in untrusted content survive summarization and get executed,…

Replication package studying how LLM vulnerability auditors defer to static-analysis evidence, with a tree-sitter C/C++ front end, controlled…

LLM safety: analyze how fine-tuning shifts refusal behavior and evaluate layer-freezing and low-rank defenses.

Open-source AI security benchmarking CLI. Measure how AI models perform offensive security tasks with MITRE ATT&CK analysis and KSM scoring.

Reproduces CVE-2026-44246, a prompt injection vulnerability in nnU-Net's GitHub Actions triage agent, demonstrating how issue content is inlined into…

Fix-Like Artifacts With Embedded Defects

A modular framework for benchmarking LLMs and agentic strategies on security challenges across HackTheBox, TryHackMe, PortSwigger Labs, Cybench,…

Production-grade MCP server giving Claude 27 security intelligence tools across 21 APIs — CVE lookup, EPSS scoring, CISA KEV, MITRE ATT&CK, Shodan,…

Artefacts for blog post on finding CVE-2025-37899 with o3

CVE-2025-23266 targets FastAPI’s parse_request() function, where oversized HTTP headers cause a buffer overflow and remote code execution. The…

This repository provides the official implementation of POISONCRAFT: Practical Poisoning of Retrieval-Augmented Generation for Large Language Models.

Security advisory detailing a critical CVE in Copilot AI where RAG-based citation links are forged to a third-party domain, enabling source…

Proof of concept for CVE-2024-24590

Collection of CVE(work) on tenserflow binary pwning it