Classify Product Launches — Data skill for Claude Code

Classify product launches: the GTM Tiering Agent

**From Victor Arellano, product marketer.** A Claude Code agent that decides how big a product launch should be, and explains why in words your product manager can use. It is **the first agent in a series of GTM agents for product marketers (PMMs).** Built and tested on **fictional data** (a made up company called Rootline), so you can try it today without any of your own company's data.

What this is, in one minute

Most companies size a launch by how loud the request was or how long the team worked on it. A big build gets a campaign and brings no leads; a small fix that closes deals gets two lines in a release note.

This agent scores each release on the roadmap against your own launch rubric and recommends a tier:

Tier	Its job
1	Tell the market
2	Tell customers and sales
3	Tell the customers it serves
4	Log it

It gathers the evidence from your sources, **asks you to score first** so it doesn't anchor you, compares its scores with yours, and writes a dated assessment with an explainer for your product manager. **You decide every tier.** The agent never approves one on its own, and it never answers for your PM.

What it writes is read by two people and by the agents that come next: the PMM, the product manager, and the next agent in the chain (brief intake), which starts the launch brief from it. The format is fixed and checked: see `5-launch-suite/1-gtm-tiering-agent/handoff.md`.

Try it in five minutes, on Rootline

1. Install Claude Code and clone this repo.
2. Open Claude Code in the suite's folder: cd classify-product-launches/5-launch-suite and run claude.
3. Type /tiering-release RL-401 (Rootline's grocer scorecard export). Any id in data/roadmap.csv works.
4. Read the evidence it shows you and score first, in one line, for example 5 4 7 5 6 8 5 · M6 · pass.
5. Compare your run with Victor's in `launches/RL-401-g