Reaching v1.0.0 on Pysealer
Artifact Development
In order to reach this v1.0.0 release of the Pysealer tool, I had to carefully plan out the basics of how the tool would work and what specific goals it would aim to solve. During the planning process, I explored existing attack vectors and security approaches for MCP servers and identified a gap that my tool could address. I spent time exploring both tool shadowing and tool poisoning attacks and demonstrated the attacks myself. After this, I determined that my skills would be suited to build a defense-in-depth tool (Pysealer) for source control that would help prevent upstream attacks by detecting code and docstring modifications. This approach is important because no single security measure can address every possible vulnerability; if one layer fails, others can still provide protection. As I reflect upon my artifict development, I believe I have come a long way since the start of the semester and have exciting plans for new artifact functionality.
Chapter One: Introduction
While building Pysealer, I simultaneously worked on the academic foundation of the project. I drafted the introduction, which provides a comprehensive overview of the Pysealer tool and its role in securing Python source code through cryptographic verification. I wanted to start the introduction with a high-level overview that explains version control concepts and how Pysealer offers a novel per-entity approach that complements traditional systems like Git. I then explored the primary motivation behind Pysealer: addressing security vulnerabilities in Anthropic’s Model Context Protocol (MCP), specifically tool poisoning and tool shadowing attacks. After this, I spoke about the current state of MCP security research, including existing benchmarking systems and security tools. Towards the end of this chapter, I clearly outlined Pysealer’s three core goals: detecting version control changes, preventing upstream attacks, and enabling defense-in-depth strategies. Finally, I addressed the ethical implications of both MCP vulnerabilities and Pysealer itself, discussing information privacy concerns and the potential for dual-use of security tools.