Over the last few months of doing Product work with AI I’ve built up a layered system of context files, prompts, slash commands, skills, and agents that I use extensively every day to review PRDs, draft OKRs, investigate customer escalations, summarize meetings, manage cross-session context, and so much more. This page collects the parts of the system I’ve made public, plus the writing that explains how it fits together.
The toolkit
Product AI Public is a curated subset of my personal “second brain” repo. It includes the prompts, context files, slash commands, skills, and agents I’ve found most useful in day-to-day product work.
A few things you’ll find inside:
- Prompts for reviewing PRDs, drafting OKRs, debating product ideas, and reviewing engineering specs against the original requirements.
- Context files that encode my product philosophy and the writing patterns I work hard to avoid (the “AI patterns” file is the one most people end up forking first).
- Slash commands for OpenCode and Claude Code —
/meeting,/code-review,/doc-review,/new-project,/session-start,/session-end, and others. - Skills that load when the topic warrants — product thinking, interactive explainers, and narrative walkthroughs of code.
- Agents for code review, blind validation of cited sources, adversarial reasoning, and ruthless style editing.
The repo is set up so each piece works on its own. You can lift one prompt or wire up the whole stack in OpenCode, Claude Code, Cursor, or whatever you use. The README walks through how to adapt the conventions to your own context.
The writing
If you want the philosophy and the evolution before the file tour, these build on each other in order:
- How I Use AI for Product Work (Dec 2025) — The foundational post. The argument is that the unlock is context plus constraints, composed for the task at hand. Covers the folder structure, the layered approach, and how I treat the LLM as a skeptical colleague.
- How My AI Product “Second Brain” Evolved (Dec 2025) — Move from Windsurf to Claude Code and OpenCode, the introduction of slash commands that remove the friction of composing context manually, and the addition of skills that apply product methodology automatically.
- How to Set Up OpenCode as Your Product Second Brain (Jan 2026) — The hands-on companion. A 30-minute setup guide that walks through creating the folder structure, configuring AGENTS.md, and building three working slash commands.
- Project Brains: Organizing Complex Initiatives (Feb 2026) — How to handle context fragmentation on long-running, cross-team projects. A dedicated folder structure with a living CONTEXT.md that tracks status, stakeholders, decisions, and open questions.
- From Assistant to Collaborator: How My AI Second Brain Grew Up (Mar 2026) — The point where the system started doing things I couldn’t easily do alone: multi-agent orchestration, cross-session memory, and domain-specific skills that encode institutional knowledge about how the products actually work.
A note on what this is for
The point of putting this in the open is that the only way to figure out what works for your role, your domain, and your team is to start with something concrete and bend it. So fork it, gut it, rename everything, throw out the parts you don’t need. If anything in here saves you a few hours, it’s done its job.
If you want to compare notes or you build something interesting on top of it, I’d love to hear about it on the contact page.