
sentrymcp
A static + runtime security scanner for MCP (Model Context Protocol) servers

A static + runtime security scanner for MCP (Model Context Protocol) servers
Securekit is a protocol-agnostic security kernel that enforces zero-trust, sandboxed execution for AI tool use. It sits between any LLM or agent…

Proof-of-concept demonstrating prompt injection in Langchain's GraphCypherQAChain leading to SQL injection in Neo4j databases. Includes Docker-based…

Research repository documenting LLM generalization ceilings in code security vulnerability detection, with cross-evaluation across synthetic and…

Benign, self-contained reproduction of CVE-2026-77521 (MaxKB prompt-injectable command execution)

Open standard for documenting security-relevant metadata of AI models, including training data provenance, PII risk, known vulnerabilities, and…

This repository contains the official implementation of the paper "[Safety in Batches? Understanding and Mitigating Safety Failures in Batch…

Lightweight evaluation harness for RAG poisoning defenses that builds four-condition clean/poisoned records and reports prevented targets, net…

A contextual security auditing system for research artifacts

A curated collection of resources for learning and researching LLM prompt injection attacks, defenses, and security.

AGLS

Research code implementing T-Backdoor, temporal-trigger backdoor attacks on spiking neural networks using rate, latency, and jitter triggers without…

Intentionally vulnerable machine learning model for hands-on security training. Explore common ML vulnerabilities, adversarial attacks, and defensive…

Weight-level LLM forensics: diff model weights, detect fine-tunes, merges and disguised derivatives, and surface tokens a fine-tune promotes or…

A complete guide and workflow for integrating Agile sprints with DevOps CI/CD pipelines.

Sandbox and MCP proxy that blocks AI coding agents from reading SSH keys, AWS credentials, and .env files, with deny-by-default policy and…

CVE-2026-20685 - Draft or TODO

Multi-stage prompt injection technique that bypasses LLM safety alignment via identity reassignment, refusal suppression, and output coercion,…