Why prompt wording actually matters this much
An AI coding tool fills in whatever you leave unspecified with an assumption — and that assumption is often reasonable but not what you meant. "Add validation to the signup form" might get you a single required-field check when you actually needed email format checking, password strength rules, and a duplicate-account check. The fix isn't a magic phrase; it's stating the actual constraints, the edge cases you care about, and what "done" looks like.
Rather than re-deriving this from scratch every time, our prompt library collects working, structured prompts by category that you can copy and adapt — the patterns below explain what makes them work.
Prompts for starting a new project
The strongest starting prompts specify the stack, the core feature, and any hard constraints upfront — rather than a single vague sentence about the end goal. "Build a note-taking app" invites the AI to guess your framework, your data storage, and your styling approach; naming these explicitly narrows the guesswork to the parts that actually require creativity.
Prompts for extending existing code
When adding to a codebase that already exists, the most useful prompts point at the specific files or patterns involved rather than describing the feature in the abstract — "add a new field to the user profile form in `ProfileForm.tsx`, following the same validation pattern already used for email" gives the AI a concrete anchor instead of asking it to invent conventions that conflict with what's already there.
Prompts for debugging
A debugging prompt is strongest when it includes the actual error message or test failure, what you expected instead, and what you've already tried — not just "this doesn't work." This is the same discipline covered in How to debug AI-generated code: specificity narrows the search space dramatically compared to a vague description of the symptom.
Prompts for tests and review
Asking an AI to write tests for code it just wrote is useful, but asking it to review that same code with a specific lens — "check this for cases where the input could be empty or malformed," "check this for any place a user could access another user's data" — tends to surface more than a generic "review this code" request, because it's pointed at a category of mistake rather than asking for an open-ended opinion.
A note on how the library is organized
The prompt library is organized by the situation you're in — starting a website or SaaS project, working in a specific framework, debugging, security review, and more — precisely so you can find a prompt shaped like your actual problem rather than starting from a blank text box every time.