Instrument

VisorAgent

Bill Rondell, Nick Adams & Dade Murphy — instrument — 1 min read — 64 words

Phase: analyze
Language: Go

Agentic LLM injection benchmark with HIT/MISS evidence packages

§ Workflow phase

01 hunt
02 analyze
03 enrich
04 report
05 instrument

Analysis. Characterises and classifies findings.

VisorAgent is an agentic LLM injection benchmark. It delivers VisorCorpus payloads to targets discovered by VisorSD and outputs a HIT/MISS matrix per attack vector with reproducible evidence packages.

The pipeline is:

VisorSD, Shodan-driven target discovery
VisorCorpus, payload generation
VisorAgent, delivery + evidence capture

Output is normalized for VisorLog ingestion so every test ends in the findings ledger.

§ analyze layer

Same phase

01

VisorCorpus

Adversarial corpus generator for LLM and RAG systems

Go
02

BARE

Air-gap-native semantic exploit mapping

Rust
03

VisorRAG

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

Go
04

Cortex

Authorization context analyzer with 14-sample reference corpus

Go