mcc_website

Imported from istiaqueahmedarik/mcc_website (AGENTS.md). Install upstream with npx skills add istiaqueahmedarik/mcc_website. Copyright stays with the author.

AGENTS.md

This repository uses an RSD-first delivery workflow. Agents must read this file before planning or editing.

Project Shape

• client/: Next.js app using React, Tailwind CSS, shadcn/Radix UI components, lucide icons, and route handlers under client/src/app/api/.
• server/: Bun-powered Hono API with PostgreSQL access under server/src/.
• Trainer/classroom UI entry points live mainly under client/src/app/trainer/, client/src/app/classroom/, and client/src/components/Navbar.js.
• Trainer/classroom API entry points live under server/src/routes/classroomRoute.ts, server/src/controllers/classroomController.ts, and trainer form routes/controllers.

Web Interface Guidelines

For any new interface design, including new pages, new product surfaces, reusable UI components, and major page redesigns, follow Vercel's Web Interface Guidelines as a quality baseline:

• Canonical source: https://vercel.com/design/guidelines
• For any new interface design, load and apply this design skill stack before choosing layout, component structure, motion, or polish details:
  • interface-design (/home/arik/.agents/skills/interface-design/SKILL.md) for product intent, domain-specific direction, hierarchy, tokens, density, states, and design-system fit.
  • apple-design (/home/arik/mcc_website/.agents/skills/apple-design/SKILL.md) for direct manipulation, fluid/interruptible motion, spatial consistency, translucent materials, typography, and reduced-motion behavior.
  • emil-design-eng (/home/arik/mcc_website/.agents/skills/emil-design-eng/SKILL.md) for component polish, animation purpose/timing/easing, press feedback, transform-origin details, and UI review format.
• Apply the current guideline unless an approved RSD records a deliberate exception.
• Before final handoff, audit generated interfaces for keyboard access, visible focus, focus management, hit targets, semantic links/buttons, URL-backed state when relevant, loading/empty/error states, responsive mobile/laptop/ultra-wide layout, long-content resilience, form behavior, reduced motion, compositor-friendly animation, contrast, and performance basics such as avoiding layout shift.
• Preserve approved project-specific design decisions, including trainer/classroom operational UI guidance, and prefer existing shadcn/Radix, Tailwind semantic tokens, and lucide icons over new abstractions.
• This rule does not approve retroactive cleanup of existing pages; use a separate approved RSD for broad redesign or compliance work.

Start Here

1. Read the user task.
2. Read the knowledge base under docs/knowledge-base/ if present.
3. Use context7 for proper documentation

Jev-Orchestrated Subagents

For substantive project tasks, load and follow .agents/skills/jev-subagent-router/SKILL.md. The main task acts as orchestrator: use Jev to select the configured model/reasoning tier and agent role, delegate the bounded work, then review and integrate the result. Keep greetings, status checks, tiny factual replies, and questions blocked on required user input in the main task.

Before presenting a multiple-choice question, use the same skill's recommendation mode to select one recommended option from the finalized choices. Put that option first and mark it as recommended only when the router reports sufficient confidence; never answer or authorize an action on the user's behalf.

Project Memory

Keep durable project memory in docs/knowledge-base/:

• project-index.md: capabilities, entry points, important files, owners
• patterns.md: reusable implementation and testing patterns
• decisions.md: durable project decisions
• mistakes.md: mistakes, near misses, and prevention notes
• quality-rules.md: repository-specific quality rules

Update project memory after every approved RSD, approved technical decision package, approved task plan, implementation review, final merge, and mistake note.

Planning Artifacts

Use these default locations:

• ADRs: docs/adr/
• Reviews: docs/reviews/

Parallel Work

Parallel agents must run in separate Git worktrees.

Rules:

• One agent per worktree.
• One branch per worktree.
• Disjoint write scopes only.
• Review diffs before merge.
• Re-run verification after integration.

Local Commands

Client:

Set-Location client
npm run lint
npm run build

Server:

Set-Location server
bun run dev

Use the narrowest verification that covers the changed files. For client navigation/UI changes, prefer npm run lint in client/, then npm run build when risk or scope justifies it.

Review Standards

Review for:

• Requirement satisfaction
• Correctness
• Maintainability
• Code-quality rules from the RSD orchestrator references
• Test coverage
• Security and privacy
• Migration and rollback safety
• Documentation and knowledge-base updates

Code Quality Bar

Agents must optimize for readable, maintainable, deliberately designed code.

Apply these rules:

• Keep designs simple and requirement-driven.
• Prefer existing local patterns over new abstractions.
• Hide volatile implementation details.
• Avoid code smells: bloaters, change preventers, dispensables, excessive coupling, and pattern misuse.
• Refactor in small test-backed steps.
• Avoid speculative abstractions and cargo-cult design patterns.
• Remove duplicate knowledge, dead code, and unclear names within the approved scope.
• Record durable quality lessons in docs/knowledge-base/quality-rules.md.

Security Checklist

Before final merge, check:

• Authorization and permissions
• Data exposure
• Input validation and injection risks
• Secret handling
• Logging of sensitive data
• Dependency risk
• Unsafe defaults

END

Write for reviewer to review:

• explain clearly what you did, why you did
• which code file change and not just random order, need to have a flow so that reviewer can review as if he wrote it