Prompt Library
Prompts that produce something you can check.
Platform-neutral prompts for scoping, building in milestones, testing and hardening. Use them with any AI build tool — the workflow is the point, not the vendor.
Browse
Filter by stage of the build
ScopingBefore you build anything.
Vertical slice scope
Act as a product lead. From the idea below, define the smallest vertical slice that a real user could complete end to end. Return: one primary user, one job to be done, the single core journey in numbered steps, what is explicitly out of scope for v1, and three acceptance criteria I can verify by hand. Idea: <describe your idea in two sentences>
BuildTurn a scope into an ordered build.
Milestone build plan
Given the scope below, produce a build plan of 5–7 milestones. Each milestone must be independently testable and end in a working state. For each, give: the goal, the files or areas affected, the acceptance criteria, and the manual test I run before moving on. Do not combine unrelated work into one milestone. Scope: <paste your scope>
BuildKeep generated work honest.
Acceptance criteria pass
Review the implementation against these acceptance criteria. For each criterion, answer PASS or FAIL with the specific evidence in the code. If FAIL, give the smallest change that fixes it. Do not add features or refactor anything outside the failing criteria. Criteria: <paste criteria>
BuildWhen the build keeps growing.
Scope-creep guard
Compare the current implementation to the agreed v1 scope. List anything present that is not in scope, and anything in scope that is missing or only partially built. Recommend what to cut before launch and what must be finished. Be blunt.
QABefore any public link.
Mobile and responsive QA
Audit this interface for mobile use at 390px wide: tap-target sizes, overflow, text truncation, sticky elements, safe areas, form usability with an on-screen keyboard, and orientation change. Return a prioritised list of concrete fixes with the element or component each applies to.
QAFind the paths nobody tested.
Failure-state sweep
List every request, form and external dependency in this feature. For each, describe what the user currently sees when it is slow, empty, unauthorised or fails, and what they should see instead. Then implement only the missing states.
QAShip something usable by everyone.
Accessibility review
Review this page for accessibility: heading order, landmark structure, label association, keyboard reachability and visible focus, colour contrast, and status messages for screen readers. Return specific violations with the fix for each. Do not change the visual design beyond what a violation requires.
ExportBefore pushing to a repository.
Pre-export hardening
Audit this project for anything unsafe to publish: hard-coded keys or tokens, privileged calls made from the browser, permissive data access rules, debug output, and unused dependencies or dead files. Return findings grouped by severity with the change needed for each.
ExportSo someone else can run it.
README for a handover
Write a README for this project covering: what it does in two sentences, prerequisites, install and run commands, required environment variables by name only, how to build, how a release reaches production, and known limitations. Keep it factual and short.
LaunchThe day before you announce.
Launch readiness check
Run a launch readiness review: unique page titles and descriptions, favicon and share image, custom domain with HTTPS, analytics, privacy and terms pages, a working contact route, error pages, and a rollback plan. Return a checklist marking each item ready or not, with the action required.