Making AI agents trustworthy, secure, one tool at a time.
Product Engineer at Corgea.
Stilgar — Security Research for AI Agent Harnesses
A security research initiative building tools for AI agent harnesses and teams. Stilgar explores how to make agent orchestration safer and more reliable at the infrastructure level.
Inconnu — Zero-Trust Bridge for RAG
RAG systems access sensitive data without proper security boundaries. Inconnu is a zero-trust security layer that sits between your RAG pipeline and your data sources.
Harness — Quality Guardrails for AI Coding Agents
AI coding agents generate code without consistent structure or quality standards. Harness provides opinionated project templates with built-in guardrails so agents produce reliable, maintainable code from the start.
Spok — Spec-Driven Development for AI Coding Agents
AI coding agents build without first agreeing on what to ship. Spok is a lightweight spec-driven workflow: explore an idea, propose a chunked change, apply one slice at a time, then fold the result back into living specs. Inspired by Dexter Horthy's talk Everything We Got Wrong About Research-Plan-Implement.
Interlocks — Deterministic Python Quality Loop
Local checks, CI, and agent-authored pull requests drift apart. Interlocks runs one deterministic Python quality loop — lint, typecheck, test, coverage, and more — behind a single command so humans and agents review against the same output.
Wando — Security Checklist for Teams
Teams lack actionable security guidance. Wando provides curated, practical security and privacy recommendations you can actually follow.