Independent Research Dossier

Security Research, Threat Intelligence, and Offensive Analysis

Threat IntelVulnerabilityOffensive ResearchWeb3 & AI Security

High-quality security research organized like a working dossier, turning complex threats and vulnerabilities into actionable knowledge for defenders and builders.

From a practical standpoint, we dive deep into attack methodologies, track threat actor behaviors, and uncover supply chain risks, aiming to build a systematic security knowledge framework.

Explore Research Latest Analysis

Featured Research

Latest Research

View all research
Over a near-black teal-blue background, a captured git bundle unpacks within intercepted network traffic as plaintext credentials and cryptographic material spill out intact and ungoverned by any redaction filter; the foreground shows a clean developer-tool facade concealing a background upload conduit dispatching the entire repository ahead of the model's inference call; the midground features a parallel conduit bypassing a security filter with raw sensitive data flowing unchecked through the ungoverned channel — copper-orange accents punctuate the scene, symbolizing the Grok CLI whole-repo upload reverse analysis
Security Research

Grok CLI Whole-Repo Upload Reverse Analysis: How a Single Prompt Ships Your .env to the Cloud

Triple cross-verified with IDA Pro, Frida, and mitmproxy, this post reproduces Grok CLI v0.2.98 uploading the entire codebase (including .env, RSA private keys, and other common dev-time secrets) verbatim and unredacted to xAI cloud storage via git bundle at the start of each turn. Compared against v0.2.93, the upload mechanism is identical across both versions, respect_gitignore defaults to false, and one of the 8 independent switches allows remote override.

2026-07-15 10 min #Sample Analysis