Analysis updated 2026-05-18
Force a product-first discussion before any code gets written for a new feature
Get a written design review automatically for changes touching money or permissions
Have technical jargon explained in plain language during planning
Separate self-verified testing from steps only the user can check
| amarrey/product-first | 00kaku/gallery-slider-block | 04amanrajj/netwatch | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | — | JavaScript | Rust |
| Last pushed | — | 2021-05-19 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | easy | easy | moderate |
| Complexity | 1/5 | 2/5 | 3/5 |
| Audience | pm founder | general | ops devops |
Figures from each repo's GitHub metadata at analysis time.
product-first is a skill for Claude Code, meant for people who build software with the help of AI but think about their project in terms of products and systems rather than code. Most workflows for building a feature jump straight into implementation. This one insists on settling what you are building and why first, including who it is for, what job it does, what success looks like, and what is intentionally left out, before any technical decisions get made. The second thing it does differently is scale how much process a change gets based on how risky it is, not how large it looks. A small copy or visual tweak just gets built right away. A change that touches money, permissions, or the database gets a written design plan and an independent review before anything is built. So a big visual redesign can stay quick and lightweight, while a small change to how permissions work gets real scrutiny. Throughout the process, technical jargon is explained in plain language, and any real fork in the road is presented as a multiple choice question with a recommendation, so someone without an engineering background can still make informed product decisions. The skill defines three tiers of process, quick for cheaply reversible changes like UI or copy, standard for typical features which get a short design summary that is approved before building, and high stakes for anything involving money, permissions, database migrations, or messages to real users, which gets a written design document and independent review. The overall flow moves from a product brainstorm, to an engineering design phase, to a written summary covering the problem, the design, what is out of scope, risks, and what counts as done, then building, then a testing hand off that separates what was verified automatically from what only the user can check themselves. It installs by cloning the repository into a Claude Code skills folder, either personally or per project, and can be invoked directly or triggered automatically when describing a feature. It is released under the MIT license.
A Claude Code skill that makes AI-assisted building start from product thinking, scaling how much design rigor a change gets to how risky it is, not how big it looks.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly pm founder.
This repo across BitVibe Labs
Verify against the repo before relying on details.