Review AI Product Idea
Run a stage-gate review on any AI product and get a Go, Conditional, or Kill decision backed by a scored readiness table, the top blocker, and the next-stage investment to release.
Install the skill or give me an instruction how to install it - ai-stage-gate from https://github.com/ML-SystemDesign/MLSystemDesign Free Open source Works in Claude, Cursor, Codex & more
What it does
You describe where your AI product is and what evidence you have. The skill locates it on a six-stage process, scores the current gate's required deliverables and transition criteria as Met, Partial, Not met, or Unknown, and reads off a traffic-light decision: Go, Conditional, or Kill.
It works from a gate deck, a repo, telemetry, or any mix, and applies dual value-and-technical validation with data, ethics, and alignment checks at every gate. LLM, RAG, and agent products get an extra modern-AI readiness row.
You get back a screenshot-ready gate scorecard with the top blocker, the conditions to advance and who owns them, and the recommended next-stage spend, followed by concrete comments and an ordered path to a clean Go.
When to use it
- You need a Go or Kill decision on an AI product
Get an evidence-based gate call, not a gut feel, with the single blocker holding it back and the conditions to clear it.
- You're defending a gate review soon
See the readiness table your committee will test and the ordered path to a clean Go before you present.
- You're triaging a backlog of AI ideas
Score several initiatives at once into one portfolio matrix and see which to fund, which to recycle, and which to kill.
- You can't tell if you're ready to advance
Locate the product on the six-stage process and find out which deliverables and criteria are still Partial or Unknown.
FAQ
How is this different from ml-system-design-review?
ml-system-design-review judges how good a design is; ai-stage-gate judges whether an AI product has earned its next investment. One grades the design, the other makes the gate decision. They are siblings in the same collection.
Does it need a formal gate deck to work?
No. ai-stage-gate grades whatever evidence you have, a gate deck, a repo, telemetry, or any mix, and marks anything missing as Unknown so the decision shows exactly what it rested on.
What decision does it return?
A traffic-light gate decision: Go, Conditional, or Kill, with stage-specific variants like Pivot, Recycle, or Limited Go. The call is computed from a readiness table with veto logic, never an average.
Is it free to use?
Yes. ai-stage-gate is open source and MIT-licensed with no paid tier. It runs locally in Claude Code and other Agent Skills consumers, against your own product evidence.