Formalizing the Safety, Security, and Functional Properties of Agentic AI Systems

This project is primarily a theoretical research paper (referenced via arXiv) with no significant code repository traction (0 stars). While it addresses a critical gap—the lack of formal security guarantees in agentic protocols like Anthropic's MCP—it lacks the 'moat' of an implementation or a developer tool. The defensibility is low because formalizing a protocol created by a frontier lab (Anthropic) is a reactive effort; if the lab updates the protocol or releases their own verification suite (which is highly likely for safety-conscious labs), this specific formalization becomes obsolete. The quantitative signals (0 stars, 183 days old) indicate that this has not transitioned from an academic exercise to a used piece of infrastructure. Furthermore, frontier labs are currently the primary drivers of agentic standards. Anthropic (MCP creators) and Microsoft (AutoGen) are better positioned to enforce safety at the protocol level. For a technical investor, the value here is in the intellectual property/insights rather than a defensible product. The 'high' frontier risk stems from the fact that safety/security is a core product pillar for OpenAI and Anthropic; they will eventually bake these formal properties directly into their SDKs.