Product Growth Team OS Starter Kit

Your team’s shared knowledge base. Built for AI-native product teams of any size.

Created by Aakash Gupta and Hannah Stulberg, based on Hannah’s implementation at DoorDash and patterns from Dave Killeen (Pendo) and Gabor Meyer (Google).

This repo extends Hannah’s open-source Team OS example repo (CC BY-NC 4.0) with shared skills, analytics templates, adoption tools, and a PM-OS migration path.

Quick Start

Small team (2-5 people)

Read MINIMAL-SETUP.md. You need 3 folders and 5 files. Start there.

Full setup (new to Team OS)

1. Edit CLAUDE.md — add your team roster, Slack channels, and product context
2. Replace the billing/ placeholder with your actual product areas across PRDs, plans, RFCs, analytics
3. Stub feature-index.yaml and analytics/data-catalog.yaml with your real features and tables (the file headers show the format)
4. Push to GitHub
5. Show one teammate three queries on Monday — see team/adoption-playbook.md

Upgrading from Aakash’s PM OS

Run /upgrade-to-team-os in Claude Code. Scaffolds the team structure around your existing setup in ~2 minutes without overwriting any personal files. See pm-os-migration.md.

What’s Inside

team-os/
├── CLAUDE.md                                    ← root routing table (loaded every session)
├── LICENSE                                      ← CC BY-NC 4.0, attributing Hannah's source repo
├── MINIMAL-SETUP.md                             ← lightweight guide for small teams
├── pm-os-migration.md                           ← guide for PM OS owners
├── .gitignore
├── .github/
│   └── pull_request_template.md
├── .claude/
│   ├── team-learnings.md                        ← accumulated patterns and PII rules
│   ├── agents/onboarding.md                     ← walks new hires through onboarding
│   ├── commands/
│   │   ├── customer-call.md                     ← orchestrator that invokes the call-summary skill
│   │   └── upgrade-to-team-os.md
│   ├── hooks/session-start.md
│   └── skills/                                  ← SKILL.md folder convention with frontmatter
│       ├── customer-call-summary/
│       │   ├── SKILL.md
│       │   └── examples/example-2026-01-27.md
│       ├── decision-log-entry/SKILL.md
│       ├── feature-launch-gate/SKILL.md
│       ├── freshness-check/SKILL.md
│       └── weekly-synthesis/SKILL.md
├── product-development/
│   ├── feature-index.yaml                       ← master feature → artifacts lookup
│   ├── CLAUDE.md
│   ├── product/
│   │   ├── PRDs/{area}/                         ← PRDs by product area
│   │   ├── customers/accounts/{customer}/       ← context, calls/summaries, calls/transcripts
│   │   ├── competitive-research/competitors/{competitor}/
│   │   ├── strategy/{roadmaps,vision,business-context}/
│   │   ├── decisions/                           ← non-architectural decisions
│   │   ├── meetings/{standup,sprint-planning,team-bi-weekly}/
│   │   ├── workflows/bi-weekly-update/          ← worked example: spec, steps, reference
│   │   ├── launch-emails/
│   │   ├── sales-enablement/
│   │   ├── processes/
│   │   └── product-context/
│   ├── engineering/
│   │   ├── plans/{area}/
│   │   ├── rfcs/{area}/
│   │   └── bug-investigations/{area}/bug-{date}-{slug}/
│   ├── data-engineering/
│   │   ├── plans/{area}/
│   │   └── rfcs/{area}/
│   ├── analytics/                               ← type-first / area-second
│   │   ├── data-catalog.yaml                    ← warehouse table registry
│   │   ├── metrics/{area}/
│   │   ├── queries/{area}/
│   │   ├── schemas/{area}/
│   │   ├── dashboards/{area}/
│   │   ├── experiments/{area}/
│   │   ├── investigations/{area}/
│   │   └── playbooks/                           ← cross-area playbooks
│   └── design/                                  ← rationale (artifacts live in Figma)
└── team/
    ├── CLAUDE.md
    ├── adoption-playbook.md
    ├── onboarding-guides/                       ← general + 6 role-specific files
    │   ├── onboarding-general.md
    │   ├── onboarding-product.md
    │   ├── onboarding-engineering.md
    │   ├── onboarding-design.md
    │   ├── onboarding-analytics.md
    │   ├── onboarding-data-engineering.md
    │   ├── onboarding-strategy-ops.md           ← non-technical contributors
    │   └── github-101.md
    └── retros/

The 5 Shared Skills

Two Registries That Carry the System

• product-development/feature-index.yaml — every feature mapped to its PRD, RFC, plan, schema, dashboard, experiments, tickets. Check this first when looking up artifacts for a specific feature.
• product-development/analytics/data-catalog.yaml — warehouse table registry: owner, refresh cadence, upstream source, used-by.

These two files are the single biggest “make AI find things in one hop” lever in the repo.

Works With

The architecture is tool-agnostic. Examples use Claude Code, but the structure works with:

• Cursor (.cursor/skills/ equivalent — same SKILL.md folder convention)
• GitHub Copilot (.github/copilot-instructions.md)
• OpenAI Codex (AGENTS.md at the same root level as CLAUDE.md)
• Any tool that reads markdown files in a repo

Contributing

Contributions welcome — bug fixes, structural improvements, new patterns. See CONTRIBUTING.md.

For security or privacy issues, see SECURITY.md.

Portability (If You Switch AI Vendors)

The repo is plain markdown and YAML in folders. Nothing about the team’s institutional knowledge depends on Anthropic, Claude, or any specific harness. If you switch from Claude Code to Cursor, OpenAI Codex, GitHub Copilot, or a future entrant: the only file that needs renaming is CLAUDE.md → the new tool’s equivalent (AGENTS.md, .cursor/..., .github/copilot-instructions.md). The skills convert in an afternoon — the SKILL.md folder pattern is the cross-tool standard. Your decisions, customer summaries, metrics, schemas, and PRDs survive untouched.

The harness is swappable. The institutional memory isn’t.

Credits and License

• Structure: Hannah Stulberg’s open-source Team OS — feature-index.yaml, data-catalog.yaml, product-area subfoldering, SKILL.md folder convention, role-specific onboarding, customer-call workflow
• Agent patterns inspired by: Gabor Meyer’s PM agents
• Living-files concept: Dave Killeen’s DEX system
• Harness engineering: OpenAI’s published methodology
• Aakash extensions: 5 shared skills (freshness-check, weekly-synthesis, feature-launch-gate, decision-log-entry, customer-call-summary), PII rules, MINIMAL-SETUP path, PM-OS migration, adoption playbook

License: CC BY-NC 4.0. See LICENSE.

Deep Dive

Read the full guide: How to Build a Team OS (Product Growth Newsletter)