Test run your PRDs.
Your code has tests. Your spec has a review meeting. Subjourney runs your PRD with simulated users and shows exactly where it breaks, before engineering, or your coding agent, builds it.
A PRD checklist can't tell you what's missing.
Checklists check that sections exist. Reviewers check that it reads well. Neither tells you whether a real user, with a real goal, could get through the feature your spec describes.
And when a coding agent builds straight from the spec, every gap becomes a guess baked into the product.
How it works.
- 01Drop in the PRD.Plus any specs or copy it depends on.
- 02Press run.Simulated users try to reach the outcome your PRD promises.
- 03Fix the paragraph.Every stall points to the line responsible, or the one that's missing.
Reading vs running.
| Reading a PRD | Running a PRD |
|---|---|
| “Looks good to me” | Three users through, one stalled at checkout |
| Edge cases if someone thinks of them | Edge-case users run every time |
| Questions arrive mid-sprint | Questions answered before kickoff |
| No way to know a fix worked | Re-run and watch it pass |
- Reading a PRD
- “Looks good to me”
- Running a PRD
- Three users through, one stalled at checkout
- Reading a PRD
- Edge cases if someone thinks of them
- Running a PRD
- Edge-case users run every time
- Reading a PRD
- Questions arrive mid-sprint
- Running a PRD
- Questions answered before kickoff
- Reading a PRD
- No way to know a fix worked
- Running a PRD
- Re-run and watch it pass
“We don't have time for another step.”
It slots in before the review meeting you already hold. Run it first, and the meeting is about decisions instead of hunting for gaps.