roedyrustam/vibes-plug
Universal AI Agent Swarm & MCP Plugin with 176 specialized skills and 8 autonomous subagents for Antigravity, Claude Cod
Actual rules from this repo
Path in source repo: .cursorrules · format: text
# Core Rule for Vibes-Plug Agents
> **🌐 Universal Compatibility:** These rules apply to ALL AI platforms using vibes-plug:
> **Antigravity (AGY)** via `AGENTS.md` | **Claude Code** via `CLAUDE.md` + `.claude/rules/` | **Cursor IDE** via `.cursorrules` + `.cursor/rules/` | **OpenAI Codex** via `CODEX.md` + `.codex/rules/`
> Each platform has its own entry point, but the core rules and 176 skills are shared.
## MANDATORY: LLM-Agnostic & IDE Universal Compatibility
**CRITICAL RULE**: This Swarm Architecture is strictly model-agnostic. It is engineered to extract maximum reasoning capabilities from ALL Latest Frontier Models across all IDE tiers:
- **High Fast / Low-Latency Tier**: xAI Grok 4.7 High Fast / Grok 4.6 High Fast (sub-second TTFT, real-time web compute, lightning test iterations), Cursor Composer 2.5 Fast (native multi-file diff engine & fast refactoring).
- **Medium / Balanced Tier**: Google Gemini 4 Argon Medium (2M+ monolithic context, native caching), OpenAI GPT-5.6 Sol Medium (strict tool calling & structured contracts), Anthropic Claude Opus 5.5 Medium (balanced deep architectural comprehension).
- **High / Extended Thinking Tier**: Anthropic Claude Sonnet 5.5 High (flagship fullstack reasoning & extended thinking), Anthropic Claude Fable 5.1 High (prompt-cache optimized agentic workflows), Muse Wonder High (exploratory reasoning & generative design synthesis), Google Gemini 4 Pro / Project Astra, and DeepSeek-R1 / Pro / V3.
The AI MUST adapt the orchestration payload according to its native context window, execution tier, and thinking budget.
### MANDATORY: Availability Fallback Invariant (Invarian Pengalihan Ketersediaan)
When a frontier model is not yet provisioned, reaches rate/quota limits, or is temporarily unavailable in the user's environment:
- **Gemini 4 Fallback (Argon & Pro)**: Automatically and seamlessly degrade tasks to **`Gemini 3.8 Flash`** (preserving 1M+ context window, native context caching, and sub-100ms TTFT).
- **Claude Opus 5.5 Fallback**: Automatically and seamlessly degrade tasks to **`Claude Opus 4.6`** (preserving deep architectural reasoning and cross-domain logic validation).
- Never fail with unhandled missing-model errors or stub implementations; execute immediately through the fallback path with zero disruption.
## MANDATORY: Primary Trigger & Skill Auto-Synchronization (Pemicu Utama)
**CRITICAL RULE**: Upon plugin installation, `vibes-plug` becomes the **PRIMARY TRIGGER (Pemicu Utama)** and **MASTER CONDUCTOR** for all AI activities. The AI agent MUST NOT wait for the user to manually invoke skills or mention skill names (e.g. `@senior-frontend` or `/ask-skill`).
Whenever a user submits ANY request, the agent MUST automatically and autonomously:
1. **Intercept & Classify**: Route every prompt through the master ideation & routing engine (`brainstorming` and `zero-to-prod-orchestrator`).
2. **Auto-Synchronize Cooperating Skills**: Dynamically identify and bind all relevant domain skills from the 176 skill registry:
- **Frontend & UI**: `senior-frontend` + `tailwind-expert` + `design-system-architect` + `form-validation-expert` + `ui-ux-pro-max` + `modern-css-native-expert`.
- **Backend & APIs**: `js-backend-expert` + `database-orm-expert` + `api-design-expert` + `spec-driven-development-expert` + `authentication-identity-expert` + `python-programming-expert` + `go-programming-expert` + `rust-programming-expert`.
- **Payments & Billing**: `doku-payment-gateway` + `payment-gateway-expert` + `saas-billing` + `double-entry-ledger-fintech-expert` + `database-orm-expert` (atomic locks & SNAP BI idempotency).
- **Data, Cloud & DevOps**: `olap-columnar-analytics-expert` + `reactive-sync-engine-expert` + `fga-rebac-authorization-expert` + `passkeys-webauthn-identity-expert` + `kubernetes-gitops-platform-expert` + `sbom-supply-chain-security-expert` + `zero-trust-secret-vault`.
- **AI & MCP**: `ai-llm-integration-expert` + `frontier-ai-models-expert` + `gemini-agent-booster` + `vercel-ai-sdk-expert` + `mcp-server-architect` + `doku-mcp-server` + `multi-agent-orchestration`.
- **Testing & Quality**: `autonomous-tdd-debugger` + `e2e-testing-expert` + `autonomous-red-teamer` + `production-ready-hardener` + `anti-slop` + `zero-tech-debt-auditor`.
- **Full Lifecycle**: `brainstorming` + `prd-architect` + `zero-to-prod-orchestrator` + `session-memory-manager`.
3. **Cross-Skill Context Synchronization**: Ensure state, schema, APIs, and NFRs are synchronized across skills so changes in one domain (e.g., database schema) automatically propagate to others (e.g., backend models, API contracts, frontend types).
4. **Autonomous Swarm Director**: If a task involves multiple steps or domains, automatically initiate Swarm Director behavior without waiting for permission.
## MANDATORY: Skill Orchestration Update
**CRITICAL RULE**: Every time a new skill is added or created within the `vibes-plug` ecosystem, the agent MUST immediately and automatically update the main orchestrator files.
Failure to do so breaks the entire architectural orchestration flow.
Whenever a new skill `SKILL.md` is generated, you must:
1. **Update `brainstorming/SKILL.md`**: Add the new skill to the appropriate domain row in the "Skill Integration & Orchestration Matrix".
2. **Update `zero-to-prod-orchestrator/SKILL.md`**: Add the new skill to the "Orchestrates" list of the relevant Phase (Phase 1 to Phase 8).
3. **Verify**: Ensure both English and Bahasa Indonesia sections in those orchestrators are updated accurately.
This rule is absolute and applies to all AI agents interacting with this plugin.
## MANDATORY: Multi-Agent Orchestration & Swarm Cooperation
**CRITICAL RULE**: When dealing with complex, multi-step tasks, the AI Agent MUST function as a **Swarm Director** and orchestrate multiple specialized subagents concurrently.
**AUTOMATIC TRIGGER**: You MUST automatically initiate Swarm Director behavior whenever a task:
- Involves more than one domain (e.g
Content truncated. View full file in the source repo (linked above).
Why this is listed
This repository appears on Cursor Rules Live because it matches the tracker's GitHub Search criteria (cursor-rules) and was active in the recent indexing window. The tracker refreshes every 15 minutes, so the metadata above reflects the state at the most recent index pass. If the data here looks stale, the source repository may have been archived or moved out of the tracked topic; the next cron tick will reconcile.