I'm building Momentum Cognition, an AI consulting practice focused on practical automation for service businesses. My background is 20+ years in web development and CTO roles, which gives me a ground-level understanding of what actually works versus what just sounds impressive.

Currently shipping Ettin (AI knowledge management - captures and organizes business knowledge by voice or text), Assay (AI readiness assessment tool), and Emma (voice AI receptionist for healthcare/service businesses). Building Open Ettin, a semantic memory system using pgvector embeddings that lets any AI tool plug into your knowledge base via MCP.

I'm active in the Upwork ecosystem and work across Claude, ChatGPT, and other AI platforms depending on what the problem needs. I'm here to connect with other founders working on AI applications, people dealing with the messy realities of productizing services, and anyone who can help me think through go-to-market strategy for AI tools in traditionally non-tech industries. Always happy to trade notes on what's actually working in AI consulting versus what's just hype.

I'm in the specification stage of building MantleOS, an agentic AI operating system that acts as executive function for personal and business intelligence — routing tasks across domain agents, managing semantic memory, and enforcing authority tiers on every action. Before I write the Agent Harness spec (the enforcement layer governing all agent behavior), I need an external security review of the architecture.

Specifically I need someone who can evaluate: a three-tier authority system designed to prevent unauthorized agent actions; an Information Flow Control / taint propagation model aligned with the FIDES framework; prompt injection and MCP security posture against OWASP AIVSS and the MAESTRO multi-agent threat model; and a scoped delegation architecture for sub-agent calls.

This is a pre-build spec review, not a code audit. The ideal advisor has experience with LLM/agent threat modeling and is familiar with at least one of MAESTRO, OWASP AIVSS, or FIDES.