Skip to content
Scroll
OblivionLabzResponse·typically within 48 hours

I build the machinery other people’s automation runs on.

Ident

Daniel J. Vermillion

dxverm·Champaign, IL

Practice

Agent infrastructure, automation, and the security discipline around it.

// 01 · THE LAB

A sovereign AI lab.Three disciplines.One operating principle: it has to work.

  • Agentic Workflow Automation & Integrations
  • Offensive Security Research
  • AI-Native Product Development

// 02 · WHAT I BUILD

Three lines of work.

PRODUCT01

Autonomous Systems

We replace repetitive human process with agents that run 24/7 and don't ask for permission.

Custom multi-agent workflows wired to your existing stack

Triggered automations across CRMs, comms, data pipelines, and APIs

Continuous operation — monitored, observable, recoverable

PRODUCT02

Adversarial Intelligence

We find what your stack leaks before someone else does — and build the tools to keep it closed.

  • Penetration testing and red-team engagements, authorized scope only
  • Purpose-built tooling for detection, response, and hardening
  • Research-grade analysis — no compliance theater

PRODUCT03

Grimoire & AI-Native Products

Intelligent software with a point of view — products that feel like they have an operating system underneath.

  • Grimoire — thirteen skills for Claude Code, shipping today
  • Custom AI-native products built around your operators, not around a demo
  • A control surface modeled on sovereignty, not settings

// 04 · HOW THE WORK RUNS

Recon to operate.

Four phases, in order, every time. You see the design before anything runs, and you keep everything that gets built.

  1. Recon

    I map the actual problem before proposing anything: the stack you already run, where the process leaks, and what the friction is really costing. You don't need a spec — producing one is part of the work.

    No scoping theater

  2. Architecture

    The system gets designed on paper first — agents, triggers, data flow, permissions, and the security surface it opens. You read it and sign off before a line of it executes.

    You approve before it runs

  3. Deployment

    It ships into your environment or mine, with the boring parts done: documented, observable, and recoverable. Every credential is yours and stays yours.

    Documented and recoverable

  4. Operate

    The system runs, and I keep it running — monitoring, iterating, and fixing what breaks. Or I hand you the keys and the runbook and step off. Both are fine; say which you want.

    Hand-off or retainer

// 05 · EVIDENCE

Four things that exist.

No testimonials, no client logos, no number I can't show you the source of. Just the work, with the address of each piece — go and check it.

  • E · 01Shipping · $99

    The Operator's Grimoire

    Thirteen skills for Claude Code, with an installer and a written guide. One-time price, current version 1.1.1.

    Skills
    13
    Version
    1.1.1
    Includes
    Installer · written guide
    Price
    $99, one-time
    grimoire.oblivionlabz.net(opens in a new tab)
  • E · 02Public · MIT

    agent-safety-policy

    A YAML allow / ask / block rules engine for coding agents. “The agent shouldn't do that” isn't enforcement; a file you can review and diff is.

    github.com/oblivionlabz/agent-safety-policy(opens in a new tab)
  • E · 03Client · in build

    Mushwellow Dry Goods

    Handcrafted tie-dye, festival pop-ups, custom textile work. Storefront and pattern library in build — shown in progress, not as a launch.

    mwdg.oblivionlabz.net(opens in a new tab)
  • E · 04Client · in build

    Josephine Leonard Photography

    Client work, in build right now — shown as in progress, not as a launch. Design, build, DNS and hosting handled rather than handed back as homework.

    jl.oblivionlabz.net(opens in a new tab)
Public repositories
Public repositories 1 · MIT
Every link above
Every link above resolves
Testimonials
Testimonials none — I won't write my own

// 06 · CORRESPONDENCE

Tell me what you need built.

A paragraph is enough: what the thing is, what it has to do, and roughly when. You don't need a specification — producing one is part of the work.

If it's a good fit I'll come back with a scope, a shape, and a price before you've committed to anything. If it isn't, I'll say so and tell you what I'd do in your position. That answer is free and it's usually faster.

What to send
What the thing is, what it has to do, and roughly when.
Response
Typically within 48 hours.
Currently
Taking work. Agent infrastructure, automation, and the security discipline around it.
No form
This is a static page and a plain address is more useful to both of us.