.cursorrules ยท diff

git:20261005.7a88132 to git:20261005.e08f000

1 added, 1 removed. Audit A to A.

# Vibes-Plug โ€” 147-Skill AI Agent Swarm Architecture (v4.2.0)
- # Universal AI Plugin for Antigravity, Claude, and Cursor
+ # Universal AI Plugin for Antigravity, Claude, Cursor, and OpenAI Codex
## Primary Trigger & Skill Auto-Synchronization (Pemicu Utama)
Upon installation, `vibes-plug` is the **PRIMARY TRIGGER (Pemicu Utama)** for all Cursor AI interactions.
- The AI agent MUST automatically intercept prompts and synchronize relevant skills without waiting for explicit `@` mentions.
- Automatically connects cooperating skills (e.g. `doku-payment-gateway` + `saas-billing` + `payment-gateway-expert`; `senior-frontend` + `tailwind-expert`).
- Multi-domain workflows trigger Swarm Topologies (Fan-Out/Fan-In, Pipeline Saga, Critic-Validator).
## About
This project uses the vibes-plug skills ecosystem with 132 specialized skills.
Each skill is a `SKILL.md` file in the `skills/` directory containing domain-specific
instructions, patterns, and best practices for the modern 2026 tech stack.
## Core Stack (2026 Edition)
- Frontend: React 19, Next.js 15, Tailwind CSS v4, TypeScript 5.8+
- Backend: Node.js 24 LTS, Bun 1.2+, Hono v4, Fastify 5, Python 3.14, Go 1.25+, Rust 2024
- Database: Prisma 6, Drizzle ORM, PostgreSQL, Supabase, Neon, Turso
- AI/LLM: Vercel AI SDK 5.x, MCP v1.9+, pgvector, RAG pipelines
## Mandatory Before Coding
1. Read the relevant `skills/<skill-name>/SKILL.md` files before writing code.
2. Think deeply before acting โ€” analyze constraints, question assumptions, validate approach.
3. For new projects, auto-generate: PRD.md, ERD.md, DOKUMENTASI.md.
## Skill Domains
- ๐Ÿค– AI & Agentic: ai-llm-integration-expert, vercel-ai-sdk-expert, deep-research-analyst, synthetic-data-finetuning-expert, pydantic-ai-expert, mcp-server-architect, vector-db-rag-expert, multi-agent-orchestration, graph-rag-knowledge-expert, local-slm-edge-ai-expert, llm-finops-router, agentic-memory-architect
- ๐ŸŽจ Design & UI/UX: design-system-architect, hig, ui-ux-pro-max, tailwind-expert, modern-css-native-expert
- ๐Ÿ–ฅ๏ธ Frontend: senior-frontend, tailwind-expert, nextjs-app-router-expert, tanstack-query-expert, vue-frontend-expert, astro-framework-expert, svelte-sveltekit-expert, solidjs-expert, angular-expert
- โš™๏ธ Backend: js-backend-expert, python-programming-expert, go-programming-expert, rust-programming-expert, api-design-expert, fullstack-expert
- โ˜๏ธ SaaS & Cloud: saas-architect, saas-billing, saas-multi-tenant, ci-cd-devops-architect, cloud-hosting-expert, payment-gateway-expert, doku-payment-gateway
- ๐Ÿ—„๏ธ Database: database-orm-expert, supabase-security-expert, data-pipeline-etl-expert
- ๐Ÿ”’ Security: authentication-identity-expert, e2e-testing-expert, production-ready-hardener, zero-trust-secret-vault, anti-slop, autonomous-tdd-debugger, zero-tech-debt-auditor
- ๐Ÿ” SEO & Telemetry: seo, data-telemetry-expert
- ๐Ÿ› ๏ธ Utilities: brainstorming, prd-architect, session-memory-manager, zero-to-prod-orchestrator, web-scraper, website-design-cloner
## Code Quality & Sovereign Anti-Slop Directive
- Zero-tolerance for AI slop (`anti-slop` skill):
* Code-first delivery with 100% complete implementations (no `// TODO`, `// ...`, or mock stubs).
* Hyper-YAGNI: avoid speculative abstractions and redundant wrapper layers.
* Meaningful comments only: explain non-obvious business/architectural WHY, never syntax mechanics.
* No swallowed errors (`catch(e){}` prohibited); structured logging (`Pino`) required.
* UI & Visual Sanitation: Enforce Linear/Stripe-caliber design. Zero generic blue-purple gradients, no neon glow orbs, no pill button addiction (`rounded-full` on rectangular elements), mandatory 5-state contract, tabular numerals (`tabular-nums`), and crisp 1px borders.
* Run `node scripts/check-anti-slop.js` to verify zero violations.
## Sovereign Token Optimization Protocol (Hemat Token)
- Enforce deterministic KV-cache prefixing via `kv-cache-prefix-optimizer` (75-90% prompt cache hit rates).
- Prune subagent context payloads: pass only targeted `CONTEXT_MAP.md` slices, not raw chat transcripts.
- Use dynamic model cascading (`adaptive-model-cascade`): `flash` for research/linting, `pro` for architecture/security.
- Ultra-compact session checkpoints (<300 tokens) via `session-memory-manager`.
- Non-Degradation Invariant: Never truncate code with placeholders or stubs to save tokens. Code must be 100% complete.
## Orchestration Flow
1. Ideation: brainstorming โ†’ prd-architect
2. Design: design-system-architect โ†’ senior-frontend โ†’ ui-ux-pro-max
3. Backend: js-backend-expert โ†’ event-driven-architect โ†’ autonomous-tdd-debugger
4. AI: ai-llm-integration-expert โ†’ mcp-server-architect โ†’ multi-agent-orchestration
5. SaaS: saas-architect โ†’ saas-billing + saas-multi-tenant + doku-payment-gateway
6. Launch: e2e-testing-expert โ†’ zero-tech-debt-auditor โ†’ seo โ†’ anti-slop โ†’ production-ready-hardener