About

About

I’m a software engineer. I started out in security research, moved on to Android and Flutter, and eventually landed on the backend. At work these days, I mostly build AI/LLM platforms, RAG, and agents.

AI makes code faster to write, but someone still has to answer the old questions. Were the requirements pinned down? Did the change break anything else? How much access should the agent get? Most of what I write here tries to sort out which work you can safely hand to AI and which decisions still need a person. I care less about showing that something runs than about whether the system is still easy to maintain down the road, and whether the next person to pick it up can make sense of it.

What I write about here

  • Coding agent skills and workflows. I put skill packs to work on concrete examples and follow them through each stage, from clarifying requirements through design and implementation to review and merge. The mattpocock/skills and pstack series each have a guide to help you pick a starting point. I also write about authoring your own Agent Skills and scaling an AI-native development process down to a personal project.
  • Security problems that agents bring in. This covers supply-chain and permission risks from third-party skills and MCP servers, and attack chains reconstructed from public evidence. These posts trace which services a tool connects to and what access it gets, and include a checklist you can work through.
  • Source-code deep dives into open-source tools. I read the source of a specific version to find out what a tool actually does and where that differs from what it claims.
  • You’ll also find notes from tech talks and a few side projects of my own.

Most posts start from a concrete question. I link to the original sources, note the version I checked whenever I can, keep verifiable facts separate from my own judgment, and spell out where that judgment applies.