Next-Gen Prompt Governance

Composable AIEngineering Governance

Deliver strict, universal engineering guidelines and specialized on-demand skills directly to Claude Code, OpenCode, Cursor, VS Code, and JetBrains. Keep your AI grounded and your context window lean.

$npm install -g rulekitx

3-Layer Memory Architecture

Delivers a universal Core floor (L1), model-selected Domain Skills (L2), and auto-detected Project Memory (L3) in perfect, structured layers.

Smart Token-Efficient Context

Only relevant rules activate when needed, cutting unnecessary context load and token usage by up to 70% in Cursor and agent CLIs.

Project Memory Autodetection

Initialize locally to scan your dependencies and generate a tailored stack summary (Next.js, Tailwind, etc.) with smart placeholders.

Built-in Intelligence

Bulletproof Prompt Orchestration

When you invoke any skill via slash command, RuleKitX seamlessly extracts the target logic and compiles the active system constraints into strict, structured XML containers.

Enforces system-level constraints first
Deduplicates rules dynamically
Separates instructions cleanly from user tasks
Composed Context
xml
<RULEKITX_SYSTEM_PROMPT version="2.0" priority="MAXIMUM">
  <MANDATORY_INSTRUCTIONS>
    You MUST follow ALL instructions in this system block...
  </MANDATORY_INSTRUCTIONS>

  <CORE_RULES priority="CRITICAL">
    # Core floor guidelines: Correctness, safety, security...
  </CORE_RULES>

  <ACTIVE_SKILLS>
    <SKILL name="architect" priority="HIGH">
      # Selected design directives, boundaries, data flows...
    </SKILL>
  </ACTIVE_SKILLS>

  <COMPLIANCE_CHECK>
    Verify you applied core floor guidelines before replying...
  </COMPLIANCE_CHECK>
</RULEKITX_SYSTEM_PROMPT>

<USER_TASK>
  build custom authentication layer
</USER_TASK>

Explore Available Skills

RuleKitX ships with 9 premium, handcrafted engineering, design, and validation directives designed to maximize LLM compliance.

/rulekitx-architectengineering

Architect

Design scalable, maintainable system architectures.

Loads alongside Core MemoryView details
/rulekitx-refactorengineering

Refactor

Improve code maintainability without changing behavior.

Loads alongside Core MemoryView details
/rulekitx-reviewerengineering

Reviewer

Review code for production risks and security.

Loads alongside Core MemoryView details
/rulekitx-safe-featureengineering

Safe Feature

Implement new features safely inside existing systems.

Loads alongside Core MemoryView details
/rulekitx-premium-uidesign

Premium UI

Build professional, accessible, production-grade UIs.

Loads alongside Core MemoryView details
/rulekitx-redesigndesign

Redesign

Improve existing interfaces without workflow disruption.

Loads alongside Core MemoryView details
/rulekitx-exec-completeexecution

Exec Complete

Produce fully working, production-ready implementations.

Loads alongside Core MemoryView details
/rulekitx-apistandards

API Standard

Design and review APIs to a consistent production standard.

Loads alongside Core MemoryView details
/rulekitx-testingstandards

Testing Standard

Write and structure tests to a consistent production standard.

Loads alongside Core MemoryView details
.cursor/rules/rulekitx-core.mdc
yaml
---
description: Always-on reasoning and safety kernel for every task. Defines decision hierarchy, verification, and safety baselines.
alwaysApply: true
---

# Core Memory (Layer 1)

The universal floor for every task. Always active, never domain-specific.

## Decision Priority
1. Correctness
2. Safety
3. Security
4. Clarity
5. Simplicity

## Warn First, Stop Only If Critical
Default to warning and proceeding with an explicit, labelled assumption.
Smart Cursor Rules

Native Layered Compliance

By initializing RuleKitX locally, the CLI generates optimal `.cursor/rules/*.mdc` rules. To prevent prompt and context bloat, rules are loaded smartly according to their architecture layer.

Core rules and Project Memory use alwaysApply: true so they are loaded everywhere. Domain skills are written as description-only so they are model-selected on-demand.

alwaysApply: true on universal core safety (Layer 1)
alwaysApply: true on auto-detected Project Memory (Layer 3)
Agent-Requested on-demand Domain Skills (Layer 2)
Create and sync your own Custom Skills