Instrument

VisorRAG

Bill Rondell, Nick Adams & Dade Murphy — instrument — 1 min read — 78 words

Phase: analyze
Language: Go

Agentic recon CLI with RAG-grounded LLM driving 6 live tools

§ Workflow phase

01 hunt
02 analyze
03 enrich
04 report
05 instrument

Analysis. Characterises and classifies findings.

VisorRAG is an agentic recon CLI: a RAG-grounded LLM drives six live tools VisorGraph, aimap, BARE, nuclei, menlohunt, OSV-scan, with every probe sandboxed in gVisor.

The LLM orchestrates the next probe based on the current evidence graph, rather than running a fixed scan-list. The result is recon that adapts to what’s actually exposed instead of running every check against every target.

Cortex provides the auth-context analyzer that gates which probes are even considered.

§ analyze layer

Same phase

01

VisorCorpus

Adversarial corpus generator for LLM and RAG systems

Go
02

VisorAgent

Agentic LLM injection benchmark with HIT/MISS evidence packages

Go
03

BARE

Air-gap-native semantic exploit mapping

Rust
04

Cortex

Authorization context analyzer with 14-sample reference corpus

Go