Zero Theory Labs
Independent Security Engineering Lab

Nothing is claimed here until a recorded run shows it.

Zero Theory Labs builds security tools for autonomous attacks and AI-assisted engineering. Our broader research suite remains in sealed range testing while our first desktop security workbench opens for early access.

Release 01 · Invite-Only Early Access · v0.4

Marlowe.

Security for AI. A little more careful.

A native desktop terminal with a cybersecurity workbench beside it. Work in real interactive shells on the left; audit codebases with 9-CLI convergence, Attack Reachability, Shannon-entropy Codebase DNA, and one-shot host-enforced approval gates on the right.

Workbench Posture · v0.4.0 macOS Validated
Local Security CLIs 9 Engines
Codebase DNA 0–8.00 b/B
Model Endpoints 4 Local · 10 Cloud
Renderer Network CSP 'none'
Semgrep · Gitleaks · Trivy · OSV · Nmap · Nuclei · TShark · Radare2 · otool
Air-gapped Ollama / LM Studio / vLLM + Claude, OpenAI, Gemini & Grok
Source-available under PolyForm Noncommercial 1.0.0
02 · In Sealed Range Testing

Additional lab tracks are currently offline.

Public pages paused during validation
Track 02 · In Lab Development

Autonomous Intrusion Containment

Kernel- and hypervisor-level isolation designed to freeze suspicious workloads mid-step before automated lateral movement completes.

Status: Sealed range evaluation · Public documentation opens after full scenario runs.
Track 03 · In Lab Development

Post-Incident Recovery Verification

Sealed replay harnessing that tests whether restored infrastructure actually holds up under real traffic rather than shallow health checks.

Status: Measurement hardening · Public write-ups return once validation completes.
03 · About the Lab

Tools built to test whether systems do what they claim.

Zero Theory Labs is an independent security engineering lab. Pages go live one tool at a time as builds are exercised and recorded. Today, early-access testing is focused on Marlowe.

Jessie Hermosillo
Principal SRE, founder

  • Three answers, never two Every check reports holds, fails, or can't tell. Unknown is never rounded up to a pass.
  • Proof before claims A capability is published only after its scenario passes on a recorded run.
  • Limits stated up front What a tool does not enforce is written right beside what it does.