Sean Lavigne builds execution authority and governance systems for AI agents, informed by a career in nuclear command and control and national-scale product delivery.

Former Nuclear Strike Advisor to the President. Senior Technical Product Manager. Open-source contributor to AI agent safety.

Background

I spent nearly a decade at U.S. Strategic Command, including time aboard the Looking Glass airborne command post advising the President on nuclear strike options. The authority control systems I operated, Permissive Action Links, didn't evaluate whether an action was wise. They verified whether it was authorized.

That experience turned out to be directly relevant to the emerging problem of AI agent safety. Agents are now executing tool calls autonomously, deleting files, writing to databases, modifying production systems. The guardrail ecosystem focuses on what LLMs say. I'm building the layer that controls what agents do.

Agent Gate is an open-source execution authority framework that intercepts agent tool calls, classifies them against policy, and enforces vault-backed rollback before destruction. SPACES is a multi-agent coordination platform I designed to solve governance, specification drift, and agent mortality in AI-assisted development.

Before this work, I spent eight years at the Defense Counterintelligence and Security Agency leading products from prototype to national scale: eApp (2M+ applications annually, zero downtime) and WatchTower ($16M+ saved, 4-year backlog eliminated). Both were built in FedRAMP/NIST-regulated environments.

Experience & Focus

  • AI Agent Safety & Execution Authority: Agent Gate, policy-driven gating, vault-backed rollback

  • Multi-Agent System Design: SPACES platform, agent coordination, context lifecycle management

  • Technical Product Management: 0→1 product delivery, roadmap ownership, cross-functional leadership

  • Regulated System Delivery: FedRAMP, NIST, DoD compliance, federal SaaS at scale

  • Data Analytics & Automation: Real-time dashboards, workflow automation, AI-assisted validation

Tools & Technologies:

  • Python, PyYAML, PostgreSQL

  • Claude Code, Anthropic API, MCP

  • n8n, vector databases (Pinecone, Supabase)

  • OPA/Rego (roadmap)

  • AWS GovCloud, ECS

  • Power BI, R, BigQuery

  • Jira, Confluence, Agile/SAFe

  • Git, GitHub