Datasets:

kondasviktor
/

vcl-ai-coding-prompts

id stringlengths 2 2	title stringlengths 16 40	category stringlengths 2 17	prompt stringlengths 156 495	works_with listlengths 1 4	difficulty stringclasses 3 values	expected_output stringlengths 21 69	tags listlengths 2 4
01	PRD from a one-sentence idea	spec	You are helping me turn a raw product idea into a buildable PRD. Idea: {{IDEA}} Target user: {{USER}} Constraint: ship an MVP in {{DAYS}} days with AI coding tools (Cursor / Claude / Codex / Gemini). Write a one-page PRD with: 1. Problem (2–3 sentences) 2. Who it is for / who it is not for 3. MVP scope (must-have use...	[ "claude", "codex", "gemini", "cursor" ]	beginner	1-page PRD with problem, users, MVP scope, non-goals, success metrics	[ "prd", "mvp", "scope" ]
02	Tech spec from a PRD	spec	Given this PRD: {{PRD}} Write a tech spec for a vibe-coded MVP: - Recommended stack (frontend, backend, DB, auth, hosting, email/payments if needed) with one sentence why each - Data entities and key fields (no full SQL yet) - Public API / serverless routes outline - Auth model - Deployment path (e.g. GitHub → Vercel...	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Tech spec: stack, data model outline, API surface, deploy path	[ "architecture", "stack", "tech-spec" ]
03	Cut scope to a 48-hour MVP	spec	Here is my feature list / PRD: {{FEATURES}} I have 48 hours. Cut ruthlessly. - KEEP for the 48h demo (max 5 items) - DEFER (with one-line why) - DROP (with one-line why) Then give a hour-by-hour build order for the KEEP list. No fluff. If something is nice-to-have, it goes to DEFER or DROP.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Cut list + keep list + 48h build order	[ "mvp", "scope", "prioritize" ]
04	Acceptance criteria for a feature	spec	Feature to ship: {{FEATURE}} Write acceptance criteria as Given / When / Then. Add: - Happy path - 4 edge cases - Failure modes and user-visible errors - Explicit out-of-scope for this ticket - Definition of done (code, tests if any, docs, deploy) Keep it short enough to paste into an issue.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Given/When/Then criteria + edge cases + out-of-scope	[ "acceptance", "qa", "definition-of-done" ]
05	User stories from a rough brief	spec	Brief: {{BRIEF}} Write 8–12 user stories in the form: As a … I want … so that … Tag each Must / Should / Could / Won't for MVP. Add a one-line note on the riskiest story.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Prioritized user stories with MoSCoW tags	[ "user-stories", "agile", "mvp" ]
06	Map this repo in 10 minutes	codebase-analysis	Explore this repository and produce an onboarding map: 1. What the product does (2 sentences) 2. Entrypoints (CLI, server, pages, workers) 3. Folder map (one line per important directory) 4. Data flow (request → storage → response) in bullets 5. Config / env vars that matter (names only, never invent secret values) 6. ...	[ "claude", "codex", "gemini", "cursor" ]	beginner	Repo map: entrypoints, folders, data flow, where to edit first	[ "onboarding", "architecture", "overview" ]
07	Find where feature X lives	codebase-analysis	I need to change: {{FEATURE}} Find: - Primary files and symbols involved - Call chain from UI/API entry to persistence - Related tests or fixtures - Likely blast radius if I edit this Do not rewrite code yet. Report only.	[ "claude", "codex", "gemini", "cursor" ]	beginner	File/symbol list with call chain for a feature	[ "navigation", "search", "feature" ]
08	Dependency and stack audit	codebase-analysis	Audit this project's dependencies and stack: - Runtime / framework versions - Notable libraries and what they are used for - Outdated or high-risk packages (based on package manifests) - Secrets handling patterns (env names only) - Suggestions: keep / replace / remove (max 7) Be evidence-based from the repo. No generi...	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Stack summary + risky deps + upgrade notes	[ "dependencies", "security", "audit" ]
09	Explain this file like I am new	codebase-analysis	Explain this file to a non-expert vibe coder: {{FILE_PATH_OR_PASTE}} Cover: - Purpose - Inputs / outputs - Side effects (network, DB, filesystem) - Gotchas - Safe places to edit vs do-not-touch Use short paragraphs and a glossary for jargon.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Plain-language walkthrough of a file	[ "explain", "onboarding", "docs" ]
10	Trace a bug from symptom to code	codebase-analysis	Symptom: {{SYMPTOM}} Environment / repro steps (if any): {{REPRO}} Investigate the codebase: 1. Ranked hypotheses (most likely first) 2. Exact files/functions to inspect for each 3. What log/assert would confirm or kill each hypothesis 4. Minimal reproduction steps if missing Do not apply a fix until I confirm the...	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Hypotheses ranked + files to inspect + reproduction steps	[ "debug", "trace", "investigation" ]
11	Debug with logs, not rewrites	debug	Bug: {{BUG}} First propose the smallest logging / assertions to locate the failure (file + what to log). After I paste results, propose a minimal fix. Rules: - Prefer small diffs - Do not refactor unrelated code - Call out any risky change	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Minimal instrumentation plan + then fix proposal	[ "logging", "debug", "minimal-diff" ]
12	Race condition / async bug hunt	debug	Suspected race / async bug: {{DESCRIPTION}} Relevant code: {{CODE}} Analyze: - Shared state and ordering assumptions - Likely race windows - How to reproduce reliably - Fix options (mutex, queue, idempotency, transactional write) with tradeoffs Recommend one fix and show the patch.	[ "claude", "codex", "gemini", "cursor" ]	advanced	Race analysis + fix options	[ "async", "race", "concurrency" ]
13	Empty response / 500 root cause	debug	Endpoint or page returning empty body / 500: {{ENDPOINT}} Error / logs: {{LOGS}} Give a systematic debug checklist for serverless / Node APIs: - Request validation - Env / secrets missing - DB connectivity - Timeout / cold start - JSON serialization Then inspect my repo for the most likely cause and propose a fix.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Checklist for API 500/empty body debugging	[ "api", "http", "serverless" ]
14	Flaky test diagnosis	debug	This test is flaky: {{TEST}} CI notes: {{CI_NOTES}} Diagnose common flake causes (time, order, network, shared state, snapshots). Propose a stabilization patch and whether to quarantine temporarily.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Why the test flakes + stabilization patch	[ "tests", "flake", "ci" ]
15	Fix without breaking callers	debug	I need to fix: {{BUG}} Constraints: - Do not break existing callers / public API - Prefer additive changes - Document any required migration Show the patch and a short "what still works" note.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Compatible fix + migration note	[ "compat", "api", "refactor" ]
16	Extract a module safely	refactor	Extract reusable logic from: {{FILE_OR_SYMBOL}} Goals: - Same behavior - Clear module boundary - Update imports / tests Propose the new file layout, then apply a minimal extraction. No drive-by renames.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Extraction plan + patch keeping behavior	[ "refactor", "modules", "structure" ]
17	Reduce duplication	refactor	Find duplicated logic related to: {{TOPIC}} Propose one shared helper (or two max). Show before/after for call sites. Do not over-abstract. Keep names boring and descriptive.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Deduped helpers with call sites updated	[ "dry", "cleanup" ]
18	Simplify a complex function	refactor	Simplify this function while preserving behavior: {{CODE}} Prefer early returns, smaller helpers, clearer names. Add a short list of what you intentionally did not change.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Simpler function with same behavior	[ "complexity", "readability" ]
19	Migrate to TypeScript types	refactor	Add sound TypeScript types for: {{AREA}} - Introduce interfaces / types where missing - Avoid `any` unless justified in a comment - Fix resulting type errors - Do not change runtime behavior unless a real bug is found (call those out)	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Typed interfaces + fixed call sites	[ "typescript", "types" ]
20	Dead code and unused exports	refactor	Scan for dead code / unused exports related to: {{AREA_OR_WHOLE_REPO}} List candidates with evidence (no importers / unreachable). Propose deletions in a safe order. Do not delete ambiguous public APIs without asking.	[ "claude", "codex", "gemini", "cursor" ]	beginner	List of removable dead code + safe deletions	[ "cleanup", "dead-code" ]
21	Write focused unit tests	tests	Write unit tests for: {{FUNCTION_OR_MODULE}} Cover happy path + 3 edge cases + 1 failure case. Match the project's existing test runner and style. Do not mock what you do not need.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Unit tests covering happy path + edges	[ "unit-tests", "jest", "vitest" ]
22	Add a regression fixture	tests	Bug we fixed / are fixing: {{BUG}} Add a regression test or fixture that would have caught it. Prefer the smallest fixture. Name it clearly after the bug.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Fixture + test that fails before fix and passes after	[ "regression", "fixture" ]
23	Smoke test checklist for a deploy	tests	Before deploying: {{CHANGE_SUMMARY}} Write a 10-step smoke checklist for staging/production. Include auth, happy path, one failure path, and a rollback signal. Optionally sketch an automated smoke script if the repo already has one.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Manual smoke checklist + optional automated smoke	[ "smoke", "qa", "deploy" ]
24	Contract test for an API route	tests	API route: {{ROUTE}} Write contract tests for status codes, response shape, and validation errors. Use the project's test stack. Keep fixtures realistic but tiny.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Request/response contract tests	[ "api", "contract", "http" ]
25	PR review: correctness first	code-review	Review this diff / PR: {{DIFF}} Structure: - Summary (2 sentences) - Blockers (must fix) - Should-fix - Nits - Questions / assumptions Focus on correctness, security, and data loss. Skip style nits unless they hide bugs.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Structured PR review with severity tags	[ "review", "pr", "correctness" ]
26	Security pass on a change	security	Security review for: {{CODE_OR_DIFF}} Check for: injection, XSS, SSRF, authz gaps, secret leakage, insecure defaults, unsafe redirects, path traversal. Report severity (Critical/High/Medium/Low) with concrete fixes. Never invent secrets; if an env var is needed, name it only.	[ "claude", "codex", "gemini", "cursor" ]	advanced	Security findings with severity and fixes	[ "security", "owasp", "secrets" ]
27	Performance hot spots	performance	Performance concern: {{SYMPTOM}} Analyze likely hotspots (N+1 queries, unbounded loops, large payloads, missing indexes, sync work on request path). Propose 3 fixes ranked by impact vs effort. Include how to measure before/after.	[ "claude", "codex", "gemini", "cursor" ]	advanced	Hotspot list + measurable fixes	[ "performance", "n+1", "caching" ]
28	Accessibility quick pass	ui	Review UI for accessibility: {{FILES_OR_URL}} Check: semantics, labels, focus, contrast (best effort), keyboard, aria misuse. List issues with severity and patch suggestions.	[ "claude", "codex", "gemini", "cursor" ]	beginner	A11y issues + patches	[ "a11y", "ui", "html" ]
29	Build a simple interactive page	ui	Build a clean, usable UI for: {{FEATURE}} Constraints: - Minimal dependencies - Works on mobile - Clear empty/loading/error states - Match existing design tokens if the repo has them; otherwise keep it simple and high-contrast Deliver files ready to drop into the project.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Minimal HTML/CSS/JS implementing the UI	[ "html", "css", "frontend" ]
30	Improve empty and error states	ui	For this screen/flow: {{SCREEN}} Write better empty, loading, and error states (copy + UI structure). Keep tone practical. Suggest one recovery CTA per error.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Copy + UI for empty/error/loading	[ "ux", "empty-state", "errors" ]
31	Form validation that helps	ui	Form: {{FORM_FIELDS}} Define validation rules, user-facing messages, and when to validate (blur/submit). Implement in the project's stack. Prevent double-submit.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Validation rules + user-facing messages + code	[ "forms", "validation", "ux" ]
32	Design a Neon/Postgres schema	database	Design a Postgres schema for: {{DOMAIN}} Include: - Tables, PKs, FKs - Useful indexes - Soft-delete vs hard-delete choice - Migration order - Example queries for the top 3 access patterns Prefer Neon-friendly serverless patterns (pooled connections, short transactions).	[ "claude", "codex", "gemini", "cursor" ]	intermediate	SQL schema with indexes and notes	[ "postgres", "neon", "schema" ]
33	Write a safe migration	database	Needed schema change: {{CHANGE}} Write a safe forward migration. Call out locks / long-running risks. Provide a rollback strategy (or explain why expand-contract is needed). Do not drop columns in the same deploy that removes reads unless safe.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Forward migration + rollback notes	[ "migration", "sql", "postgres" ]
34	Fix an N+1 query	database	Suspected N+1: {{CODE_OR_DESCRIPTION}} Show the current query pattern, why it N+1s, and a fixed version (join, IN list, or dataloader-style batch). Include an index recommendation if needed.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Batched/joined query patch	[ "n+1", "sql", "orm" ]
35	Idempotent writes and unique constraints	database	We need idempotent processing for: {{EVENT_OR_WEBHOOK}} Design: - Idempotency key - Unique constraints - Upsert vs insert-ignore - Retry behavior Show SQL + application code sketch matching the repo.	[ "claude", "codex", "gemini", "cursor" ]	advanced	Idempotency design + SQL/code	[ "idempotency", "constraints", "webhooks" ]
36	Design a serverless API route	api	Design API route: {{ROUTE_PURPOSE}} Specify: - Method, path, auth - Request/response JSON schemas - Status codes - Rate-limit / abuse notes - Handler sketch matching this repo's style (e.g. Vercel serverless) Keep cold-start friendly.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Route contract + handler sketch + errors	[ "api", "serverless", "vercel" ]
37	Webhook handler that won't double-charge	api	Implement / harden a webhook handler for: {{PROVIDER}} (e.g. Stripe) Requirements: - Verify signatures - Idempotent fulfillment - Clear logging without leaking secrets - Failure modes that are safe to retry Align with existing code if present. Never hardcode secrets.	[ "claude", "codex", "gemini", "cursor" ]	advanced	Verified webhook handler with idempotent fulfill	[ "webhooks", "stripe", "idempotency" ]
38	Input validation for public endpoints	api	Public endpoint: {{ENDPOINT}} Add strict input validation (prefer the repo's schema lib if any). Return safe 400s. Reject unexpected fields if appropriate. Show tests for bad payloads.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Schema validation + safe error responses	[ "validation", "zod", "security" ]
39	Paginate a list endpoint	api	Paginate: {{LIST_ENDPOINT}} Prefer cursor pagination if the dataset can grow; otherwise offset with clear limits. Document max page size. Update clients/examples if in-repo.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Cursor or offset pagination implementation	[ "pagination", "api" ]
40	README that gets someone running	docs	Rewrite or create README for this project: - What it is - Quick start (copy-paste commands) - Required env vars (names + purpose, no values) - Scripts - How to deploy - License / contribution note Keep it short. Prefer accuracy over marketing.	[ "claude", "codex", "gemini", "cursor" ]	beginner	README sections: what/why/quickstart/env	[ "readme", "docs", "onboarding" ]
41	Deploy checklist for Vercel/GitHub	deploy	Change ready to ship: {{CHANGE}} Produce a deploy checklist for GitHub → Vercel (or this repo's host): - Env vars to verify - Migrations - Smoke tests - Monitoring to watch - Rollback steps Be specific to this repo when possible.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Pre-deploy checklist + rollback	[ "vercel", "github", "deploy" ]
42	Write project agent rules	agent-rules	Create a project rules file for AI coding agents working in this repo. Include: 1. Product purpose (2 sentences) 2. Stack facts 3. Hard constraints (do not invent secrets, do not commit .env, prefer small diffs) 4. Code style pointers matching the repo 5. Where docs/tests live 6. Commands to run before claiming done ...	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Rules file body ready to save as AGENTS.md / .cursorrules / etc.	[ "agents", "cursorrules", "agents-md" ]
43	Cursor: scoped agent task	agent-rules	Write a Cursor agent task prompt for: {{TASK}} Include: - Goal - Files/directories in scope (and out of scope) - Constraints - Definition of done - What to run to verify Keep the agent from wandering the whole monorepo.	[ "cursor" ]	beginner	Cursor agent prompt with files-in-scope and done criteria	[ "cursor", "agent", "scoped" ]
44	Claude Code: plan then edit	agent-rules	Task: {{TASK}} Write a Claude Code / Claude prompt that: 1. First produces a short plan (files to touch, risks) 2. Waits for my approval (tell me to reply APPROVE) 3. Only then implements with minimal diffs 4. Ends with a verification checklist Emphasize reading before writing.	[ "claude" ]	beginner	Two-phase Claude prompt: plan approval then implement	[ "claude-code", "plan" ]
45	Codex/Gemini CLI: non-interactive fix	agent-rules	I will run this in Codex or Gemini CLI non-interactively. Task: {{TASK}} Write a single prompt that includes context, constraints, exact deliverables, and shell commands to verify. Assume the agent can edit files and run commands. Fail closed if verification fails.	[ "codex", "gemini" ]	intermediate	Self-contained CLI agent prompt with verify commands	[ "cli", "codex", "gemini" ]
46	Outline an Apify Actor idea	apify-teaser	Help me shape a public-data Apify Actor idea. Niche: {{NICHE}} Produce: - Buyer sentence (who pays, for what) - Input fields (5–8) - Output dataset columns - Why public data / compliance boundary (what we will not scrape) - MVP build steps (8 bullets) - How to validate demand before coding Keep it practical. Point ...	[ "claude", "codex", "gemini", "cursor" ]	beginner	Actor wedge: buyer, input, output, compliance note	[ "apify", "actors", "ideas" ]
47	Draft Actor input schema fields	apify-teaser	Actor purpose: {{PURPOSE}} Draft input schema fields: name, type, required, default, description, editor hint. Include start URLs or search mode, max items, proxy toggle placeholder, and a smoke-test note. Output as a table then as JSON-shaped sketch for Apify input schema. Do not invent proprietary templates from pa...	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Input schema field list with types and defaults	[ "apify", "input-schema", "crawlee" ]
48	Actor README that sells the job	apify-teaser	Write an Apify Store README outline for: {{ACTOR_NAME_AND_JOB}} Sections: - What it does (job-first) - Example input - Example output rows - Pricing note placeholder - Limitations / compliance - Related Actors ideas (2) Tone: clear and honest. No hype.	[ "claude", "codex", "gemini", "cursor" ]	beginner	Store-ready README outline with example input/output	[ "apify", "readme", "store" ]
49	Cheerio vs Playwright decision	apify-teaser	Target site characteristics: {{SITE_NOTES}} Should I use Cheerio (HTTP) or Playwright for an Apify/Crawlee Actor? Decide with rationale, cost/speed notes, and a minimal approach for MVP. If unknown, say what to probe first.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Decision with rationale and starter approach	[ "apify", "crawlee", "playwright", "cheerio" ]
50	PPE pricing sketch for an Actor	apify-teaser	Actor: {{ACTOR}} Cost drivers I know: {{COSTS}} Sketch a pay-per-event pricing model: - Billable events - Suggested price bands (ranges, not fake precision) - Free-tier / smoke run advice - Questions I still need to answer before publishing Be honest about uncertainty. Link thinking to Apify PPE concepts generally.	[ "claude", "codex", "gemini", "cursor" ]	intermediate	Pay-per-event sketch with unit economics questions	[ "apify", "pricing", "ppe" ]

VCL AI Coding Power Prompts

50 battle-tested prompts for Claude Code, Codex, Gemini CLI, and Cursor — by Vibe Coder's Life.

Free catalog for vibe coders. Replace {{PLACEHOLDERS}} with your facts. Not the paid Apify Playbook prompt pack (those stay private).

Load

from datasets import load_dataset

ds = load_dataset("kondasviktor/vcl-ai-coding-prompts", "prompts")
print(ds["train"][0]["title"])

Columns

Column	Description
`id`	Stable id (`01`–`50`)
`title`	Short name
`category`	e.g. `spec`, `debug`, `apify-teaser`
`prompt`	Full prompt text with `{{PLACEHOLDERS}}`
`works_with`	Tools: `claude`, `codex`, `gemini`, `cursor`
`difficulty`	`beginner` / `intermediate` / `advanced`
`expected_output`	What a good run should produce
`tags`	Search tags