10 production grade promptova koje možeš copy-pastati

Kad prosječan korisnik otvori Claude ili ChatGPT, napiše nešto u stilu: "Izradi web stranicu za moju agenciju" ili "Napiši mi kod za funkciju pretrage". To je ekvivalent zadavanja posla junior developeru koji je jučer započeo praksu — dobit ćete najgeneričnije rješenje, krcato tehničkim pretpostavkama koje se mogu pokazati potpuno promašenima za vaš stvarni use case.

Jer, kada tako komuniciraš sa AI-jem, koristiš ga kao juniora.

Claude, a napose njegovi Opus i Sonnet modeli (te OpenAI-jev GPT-4o) najbolje performaju kada im u promptu definiraš inženjerski standard. Morate im osigurati okvir u kojem će razmišljati. Svaki kvalitetan "production grade" prompt mora sadržavati pet osnovnih stupova:

• Ulogu (Role): Tko je AI u ovom trenutku? Koji je njegov level senioriteta?
• Ograničenja (Constraints): Što *ne smije* napraviti? Koje biblioteke su zabranjene? Koji je limit *dependencyja*?
• Arhitektonska očekivanja (Architecture): Kakvu logiku slijedi? SSR (Server-Side Rendering), SPA, mikroservisi?
• Output format: Očekuješ li samo JSON? Ili kompletno objašnjenje s Markdown komentarima, Mermaid dijagramima?
• Real world kontekst (Context): Koja je poslovna vrijednost featurea? Kome je namijenjen? Tko su krajnji korisnici?

Kada spojiš ovih pet elemenata, rezultati prestaju biti "igračka" i postaju kod spreman za produkciju.

U ovom tekstu pripremio sam ti mojih 10 najdražih copy-paste promptova koje svakodnevno koristim za planiranje, pisanje, refaktoriranje i debuggiranje koda na enterprise projektima.

1. Production Feature Builder

Ovaj prompt koristim na samom početku izgradnje nekog odvojenog featurea. Njegova je moć u tome što tjera model da prvo planira (razmisli o rubnim slučajevima i arhitekturi), a tek onda ispisuje linije koda.

Prompt koji unosite: > Act as a senior staff software engineer responsible for shipping production-ready features. Your goal is to design and implement a scalable, maintainable feature with clean architecture. Before writing any code, analyze requirements, identify edge cases, define architecture, and plan implementation. Then build step-by-step. > > Feature: [Opiši feature, npr. Blog kategorizacija i paywall] > Users: [Tko su korisnici i koliko ih ima, npr. 10 000 mjesečnih čitatelja] > Tech stack: [Tvoj stack, npr. Next.js 15, Tailwind, TS, Web3Forms] > Constraints: [Npr. no extra dependencies, SSR only, minimal client boundaries] > > Your output must include architecture overview, folder structure, data flow, full implementation code, and tests outline.

2. Codebase Understanding & Refactor

Kada uskočiš u ogroman repozitorij u čijem kreiranju nisi sudjelovao, prva instinktivna reakcija je glavobolja. Ovaj prompt tjera AI da analizira spagetti kod, pronađe dupliciranu logiku, bottlenecks i sigurnosne aspekte prije nego započne refaktoring.

Prompt koji unosite: > Act as a senior engineer onboarding into a large unfamiliar codebase. First understand the architecture and data flow. Then identify structural issues, duplicated logic, performance bottlenecks, and maintainability risks. After that, propose improvements and provide refactored code. > > Codebase: > [Zalijepi sirov kod ili dokumente komponente] > > Return architecture overview, problem areas, refactor strategy, improved architecture, and rewritten clean code. Keep functionality identical while improving quality.

3. Senior Debugging Engineer

Bug u produkciji? Aplikacija puca na neočekivanim mjestima? Umjesto klasičnog "popravi mi kod", želiš da AI prvo simulira uvijete pod kojim kod puca.

Prompt koji unosite: > Act as a senior debugging engineer investigating a production bug. Carefully analyze the code, reason step-by-step, identify the underlying root cause, and propose a robust fix. Consider edge cases and performance implications. Do not skip any steps in the execution trace. > > Bug description: [Što se dešava?] > Code: [Tvoj kod sa greškom] > Error log: [Traceback greške] > > Explain what the code does, what precisely is wrong, why it acts that way, edge cases to be mindful of, and provide the fixed production-ready code.

Dobro, pokrili smo osnove za postavljanje temelja i debuggiranje. Sada dolazimo do onih naprednijih stvari. Sljedećih 7 promptova transformirat će arhitekturu tvojih projekata i omogućiti ti da iz temelja optimiziraš performanse modela baze, osiguraš komponente te uspostaviš state-of-the-art deployment pipelineove.

4. Full App From Scratch (MVP Architect)

Ovaj prompt koristim kad tek startam projekt. Cijela poanta ovdje nije napraviti "Hello world" u Next.js-u već stvoriti skalabilan kostur aplikacije koji neće trebati prepravljati za mjesec dana.

Act as a senior full-stack engineer building a complete production-ready app. First design the system architecture, then implement a minimal but scalable version. Think deeply about database design, API structure, UI architecture, and global state management before even touching the code. > > App idea: [Tvoja ideja, npr. SaaS za praćenje kalorija AI prepoznavanjem slika] > Core features: [Feature 1, Feature 2] > Users: [Kome prodajemo?] > Tech stack: [Supabase, Next.js, Stripe, Resend...] > > Return entirely detailed architecture, comprehensive folder structure, database SQL schema, standardized API routes, responsive UI breakdown, and the foundational full code. Design this strictly like a real startup MVP engineered to scale organically.

5. Security & Threat Modeler

Većina developera zaboravi na sigurnost dok im se ne dogodi incident. Ovaj prompt tretira se kao tvoj virtualni Pen-tester. Tjera AI da kod pregleda iz perspektive napadača, tražeći propuste poput SQL injectiona, XSS-a, CSRF-a i loše konfiguriranog CORS-a.

Act as a lead cybersecurity architect and penetration tester. Review the provided code systematically in order to identify potential security vulnerabilities, logical flaws in authentication/authorization, and injection vectors. > > Code to analyze: > [Vaš osjetljivi kod komponenti] > Context: [Gdje se navedeni kod izvršava?] > > For each finding: describe the exact vulnerability, determine its severity (Low/Mid/High/Critical), explain how a malicious actor would exploit it, and provide the patched, secure version of the code implementing industry standard best practices.

6. Performance Auditor & Optimizer

Imate bottleneck u loadanju stranice? React komponente gube po 30ms na besmislene `useEffect` i re-tendere? S ovim promptom tjerate model da analizira Critical Rendering Path i memoy leaks.

Act as a Principal Frontend/Backend Performance Engineer. Analyze the code specifically looking for time and space complexities, memory leaks, unnecessary re-renders, N+1 query problems, and network-level bottlenecks. > > Target Code: [Vaš kod] > > Identify inefficiencies, explain why they occur under heavy load, and provide heavily optimized code. Suggest any caching layers (Redis), indexing strategies, or memoization (useMemo/useCallback) rules that should be strictly applied here.

7. The Ultimate UI/UX Revamper

Frontend često izgleda dobro sve dok ga korisnik ne isproba. Ovaj prompt se fokusira na usability, accessibility (a11y), semantički HTML i mikrointerakcije koje razlikuju amaterske projekte od premium SaaS aplikacija.

Act as a Senior UX/UI Engineer and accessibility (a11y) specialist. I will provide you with a functional interface component. Your job is to elevate it to a premium, state-of-the-art user experience. > > UI Component Code: [Tvoj React/Vue kod] > > Refine the visual hierarchy, implement dynamic smooth transitions (framer-motion or native CSS), ensure flawless responsive design across all edge-case viewports, and guarantee strict WCAG 2.1 AA compliance (ARIA attributes, keyboard navigation). Provide the completely redesigned, polished component.

8. State Management Strategist

Kada prop-drilling postane vaša noćna mora, vrijeme je za strukturalne odluke.

Act as a Senior React/Frontend Architect specializing in complex global state management. I am struggling with convoluted state distribution and prop-drilling. > > Current implementation patterns: [Opiši/Zalijepi kako trenutno prenosiš podatke] > > Evaluate the current flow and recommend an appropriate architectural pivot (Zustand, Redux Toolkit, Context API, React Query/SWR). Justify your selection practically, then implement a centralized store solution with strictly typed state selectors and reducers.

9. Comprehensive Test Suite Architect

Svi pričaju o testovima, nitko ih ne piše dok produkcija ne izgori. Uz AI generiranje testova više nema izgovora.

Act as a Lead SDET (Software Development Engineer in Test). I need to secure 95%+ test coverage on a highly critical business component. > > Component to Test: [Vaša logika/komponenta] > Testing Frameworks: [Jest, React Testing Library, Playwright/Cypress] > > Output a highly robust, isolated suite of unit tests evaluating pure functions, coupled with integration tests validating asynchronous edge cases. Include structured mocking for any external dependencies or API endpoints. Write robust, non-brittle selectors.

10. API Contract & Middleware Orchestrator

Dizajniranje API endpointova koji nemaju smisla jedan s drugim dovodi do kaosa na klijentu.

Act as an Enterprise Backend Systems Architect designing REST/GraphQL interfaces. I need a robust API design and tightly structured middleware pipeline for managing authentication, rate limiting, and request validation. > > Requirements: [Npr. auth flow s JWT, 10 upita u minuti, validacija sheme] > > Provide a comprehensive API contract specification (endpoints, headers, standardized JSON response bodies for both success and error payloads). Follow this up with concrete code implementation for the middleware interceptors to enforce the aforementioned constraints.

--- Tu imaš 10 esencijalnih oružja u arsenalu svakog modernog developera. Integriraj ih u svoj dnevni workflow spremanjem u Notion tablicu ili TextExpander i gledaj kako ti se vrijeme razvoja prepolovljuje, a kvaliteta isporučenog koda utrostručuje!