Ayush grit kaushik

Noteer is an Autonomous AI Security Fabric designed for modern agentic workflows (Devin-style pipelines, Claude Engineer, VS Code copilots). Modern coding agents generate thousands of lines of unverified code, introducing severe risks like hallucinated package names (Slopsquatting RCE), prompt-injected exfiltration sinks, and API reverse-engineering. Traditional linters only scan code after it is written to disk, leaving a critical execution window where malicious postinstall hooks or malicious dependencies can execute. Noteer introduces a zero-trust Dual-Pillar architecture: 1. Pillar I (Pre-Write CST Gatekeeper): Hooks into editor buffers and intercepts LLM tokens in flight before disk commit. It parses code across 10 languages (Python, Rust, TS/JS, Go, C++, etc.) using high-performance Tree-sitter Concrete Syntax Trees (CST), tracing taint sinks and blocking hallucinated dependencies in under 4ms. 2. Pillar II (Runtime API Clone Shield & Cognitive Arbiter): Defends proprietary LLM and agent APIs using DPoP (RFC 9449) cryptographic token binding, preventing MITM replay attacks and reverse-engineering of agent reasoning pipelines.