Security Achievements
Live security work from the Executive Mind Pack — privacy-first tooling, local-first architectures, and verified findings. Each entry documents what we built, how we tested it, and what it does in production. This page grows as we ship new work.
August 25, 2026
Running OpenAnt Fully Local: The Ollama Adapter
We built a first-class Ollama adapter for OpenAnt — an open-source LLM-assisted SAST scanner — so vulnerability scans run against local models. No cloud, no API keys, no source code leaving the machine. Verified live: both planted SQL injection findings (CWE-89) detected via local mistral-small3.2 running on commodity hardware. Upstream PR pending.
Next achievement · in progress
This is a working corner. Each milestone we ship against the Cyber Navigators BD pipeline — and every privacy-first capability we land for client engagements — drops in here.
Why this page exists
Most "security" pages on AI consultancies are marketing copy. This one is a build log. We only publish work we've actually shipped and verified against real test corpora — with the failure modes documented, not airbrushed out.
If you're evaluating us for a security-sensitive engagement — air-gapped infrastructure, regulated data, or code that simply shouldn't leave your perimeter — get in touch. This is the work we do.