How I work with agentic AI
This is a living, inspectable snapshot of how I actually work with AI day to day — not a “look what I did” demo, but the working system underneath, kept public and versioned so you can see it and watch it change.
It’s all in one place: github.com/kn0wsnothing/agentic-ai-skills-and-tools — the skills I run, the prompts I use, a guide to how it fits together, and a changelog of what’s moved.
The shape of it
Three roles and one substrate:
- A thinking/prose agent I drive for writing, reviewing, and planning.
- A data agent that owns analysis and structured work, kept separate on purpose.
- A shared, version-controlled memory both of them read and write — plain text, so it’s inspectable and it compounds.
- Skills — structured, portable workflows — running on top: reviews, session wrap-ups, prose scrubs, a writing pipeline.
I describe the whole thing in roles, not products. My particular tools are one instantiation; the patterns are the point, and they port to whatever you use. That’s the bet: build the system around durable roles and a shared-memory contract, and swap tools underneath as they change.
Start here
- The guide — the system, the principles underneath it, and one real workflow walked end to end.
- The skills — the workflows I actually run, each portable to any agent.
- The prompts — the role/persona prompts I use.
- The changelog — what’s changed, tagged by month.
I update this as the system evolves — it’s part of my monthly review. The newsletter is where I write about the thinking behind it.