npx agentspec-cli install c4a9dcf3-ba69-4e11-b941-b301a5089285 --file-type cursorrules→ .cursor/rules/
# FirstQA – Cursor Rules ## Project **Name:** FirstQA **Role:** AI-powered QA Engineer – the first QA hire for startups. **What FirstQA does end-to-end:** - Analyzes requirements, design, and code - Creates test recipes from PRs and tickets - Automates test recipes (Playwright) - Performs AI code review - Executes tests in the browser like a manual tester - Shares test results and release decisions - Integrates with GitHub, Linear, Jira --- ## Tech Stack | Layer | Technologies | |-------|--------------| | Backend | Node.js, Express | | Frontend | EJS views, Bootstrap 5 | | Database | Supabase (PostgreSQL) | | Auth | Supabase Auth, GitHub OAuth, Linear OAuth, Jira OAuth | | AI | OpenAI API | | Testing | Playwright | | Integrations | GitHub App, Linear API, Jira (Atlassian Connect) | | Payments | Stripe | | Other | nodemailer, Axios | --- ## Project Structure ``` backend/ # Server-side ├── routes/ # Express routes ├── services/ # Business logic ├── utils/ # Utilities ├── lib/ # Shared libs └── ai/ # OpenAI client, prompts frontend/ # Client-side ├── views/ # EJS templates ├── public/ # CSS, images, logos └── chrome_extension/ docs/ # plan.md, features.md, marketing/, customer-support/ scripts/ # Dev/ops scripts supabase/ # DB migrations webhook-server.js # Entry point ``` --- ## Code Rules - Plan before implementing; explain the approach first - Use ES6+ JavaScript (async/await, no callbacks) - Use clear variable names; avoid abbreviations - Add error handling for all async work - Add Playwright tests for critical user flows - Prefer `@specific-files` over `@codebase` for context --- ## CRITICAL – Behavior ### Only act when requested - Do not change code when the user is asking questions or investigating - Do not assume the user wants code changes unless they say so - Do not refactor, restyle, or “improve” without being asked ### When to make changes - User clearly asks: “Fix X”, “Add Y”, “Implement Z”, “Change A to B” ### When not to make changes - User asks: “Why is X broken?”, “How does Y work?”, “What’s wrong with Z?” ### If unsure - Ask for clarification instead of guessing ### Before changing code - Investigate the issue - Summarize findings - Propose an approach and wait for confirmation if the change is non-trivial ### UI / styling - Do not change colors, fonts, or layout unless explicitly requested --- ## Security - Use environment variables for secrets (no hardcoded keys) - Validate and sanitize inputs - Follow least-privilege access --- ## Cost Control - Default to Auto mode - Use specific file references (`@file`) instead of broad codebase search - Plan work before running tools
Cursor AI Rules - Next.js Performance & Best Practices
// ✅ GOOD: Server Component fetching data export default async function Page() { const data = await fetchData(); return <ClientForm data={data} />; // ❌ BAD: Client Component fetching data export default function Page() {
by AgentHub Crawler
Cursor AI Development Rules
Cursor IDE rules and conventions for AI-assisted development.
by AgentHub Crawler
🛠️ Principal QA Engineer – Cursor Rule
Become a senior-level QA engineer who: > *MCP = “Managed/Media/Message Control Platform” (adapt acronym to project context).*
by AgentHub Crawler
Sign in to leave a comment.
Cursor AI Assistant Rules
1. Topic identification 2. Technology stack analysis 3. Context assessment 5. Multiple approach evaluation 6. Option comparison and trade-offs
by AgentHub Crawler