←Back/Product/Codex/Claude
IntermediateProduct

Build a Personal PM Operating System with a Shared Knowledge Repo

Create a repository-based system to provide an AI assistant with durable context about your team, projects, and communication channels. This enables it to perform routine tasks like summarizing daily updates and identifying stakeholders, acting as a personalized chief of staff.

Illustration of Build a Personal PM Operating System with a Shared Knowledge Repo

It's a really, really basic repo. It's got an agents.md, which just tells it how to behave. It's got a bunch of related skills that tell it when to fire and what to do. It's got a couple of command hooks... then their workspace is a sub repo so that if you wanna write stuff back, you can write it back privately.

Before you start

What you need

  • A Git repository.
  • Markdown files containing context about your role, team, and projects.
  • An AI assistant capable of reading local files (e.g., Codex, or a terminal setup).

What you’ll make

An AI assistant that can perform personalized tasks, such as summarizing your key Slack channels ('What's Up') or identifying the correct stakeholders for a document review, without needing to be re-prompted with this context in every session.

Tools used

  • Codex

    OpenAI's cloud-based AI software engineering agent that can execute code, run tests, and handle complex multi-file tasks autonomously.

    VisitCodex
  • Claude

    Anthropic AI assistant

    VisitClaude

The workflow

Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.

5 steps

Step 01

Initialize Your PM OS Repo

Create a basic Git repository. This will be the foundation of your OS. Inside, create a file named agents.md to define the assistant's core persona.

Example prompt
This is an example for `agents.md`: "You are my chief of staff. You know my work, my team, and my priorities. Your goal is to help me be lazy in the right places by handling routine coordination and summarization tasks."

Using a Git repo allows for version control and sharing of the core OS structure while keeping personal context private.

Step 02

Create a Memory File

Create a file that stores essential context about your work. This includes your name, your role, your manager, and the names of your primary engineering, design, and support partners.

This file is the source of truth for stakeholder management. Use it to suggest stakeholders, then verify the list before sending anything.

Step 03

Develop a 'What's Up' Skill

Create a new file in a 'skills' folder that defines a daily briefing skill. In this file, list the specific Slack channels and project boards that are most important to you.

Example prompt
This is an example prompt for the skill file: "When I ask 'What's up?', you will read the latest messages in these Slack channels: [#team-channel, #project-alpha, #growth-pms]. For each channel, read the entire thread to understand the full context before summarizing any action items or open questions for me."

Instructing the agent to read whole threads is crucial for it to distinguish a new request from one that was resolved later in the conversation.

Step 04

Structure for Privacy and Collaboration

Set up your repository using a parent/sub-repo structure. The parent repo holds the shared OS template and skills, while each PM has their own private sub-repo for their personal 'workspace' and 'memory' files.

Figma used a shared template and per-person workspaces. Check repository permissions and submodule configuration to keep private context private.

Step 05

Run Your Skills Daily

Integrate the repo with your AI assistant (e.g., by pointing Codex to the local folder). Start your day by running your 'What's Up' skill to get a tailored briefing.

Ezra Mechaber reports running this skill about six times a day to stay on top of projects and communications without living in Slack.

What good looks like

  • The assistant correctly identifies your key collaborators when asked.
  • The 'What's Up' skill provides a relevant and accurate summary of your specified projects and Slack channels.
  • The system reduces the repetitive work of providing the same context in multiple prompts.

Build your next product with ChatPRD

Turn an idea into a PRD, user stories, and a plan.

Try ChatPRD free

Runbook notes

How to recover when the loop fails and where human judgment helps.

If it goes sideways

The assistant provides outdated information because the context files in the repo are stale.
Periodically review and update the context files, especially when team structures or project priorities change. Treat the repo as a living document, not a one-time setup.
As underlying AI models improve, some detailed instructions in the skills become overly specific and may hinder performance.
Regularly prune instructions that are no longer necessary. Test if the model can achieve the same result with a simpler prompt, and remove instructions that don't 'earn their place'.

Start shipping better products.

Join 250,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.

Free to start · 14-day Pro trial · SOC 2 Type II