Poldarn

A Claude Code plugin that enforces code quality principles, provides deep investigation capabilities, and detects hidden technical debt.

Features

Skills

Skill	Description
Code Quality Principles	Architectural guidance for single responsibility, modular design, type-first development, functional programming style, and pragmatic library adoption
Deep Investigation	Systematic research methodology for resolving issues where LLM training data may be outdated
Fake Implementation Detection	Detection patterns for behavioral mimicry - code that appears complete but performs no real work
TypeScript Type Scanner	Guidance on TypeScript error classification and type health patterns
Vue Query Patterns	Best practices for @tanstack/vue-query in Vue 3 applications

Agents

Agent	Description
code-quality-review	Performs code quality review focusing on architectural principles and functional programming style
codex-reviewer	Deep code review with Codex assistance for functional correctness and regression detection
deep-investigation	Exhaustive external research for library/framework issues where training data may be outdated
fake-implementation-detector	Scans codebases for hollow implementations and placeholder code
typescript-type-scanner	Scans TypeScript projects and generates categorized health reports

Commands

Command	Description
/poldarn:ts-scan	Scan TypeScript project for type errors and generate categorized health report

Installation

From GitHub

claude plugin marketplace add PoldarnKJ/poldarn-claude-plugin
claude plugin install poldarn

Manual Installation

git clone https://github.com/PoldarnKJ/poldarn-claude-plugin.git
claude plugin marketplace add ./poldarn-claude-plugin
claude plugin install poldarn

Usage

Code Quality Review

The plugin automatically provides architectural guidance when implementing features:

User: "Implement a user authentication module"
Claude: [Applies code quality principles: SRP, modular architecture, type-first development]

TypeScript Health Scan

User: "/poldarn:ts-scan"
Claude: [Runs type check, categorizes errors, generates health report]

Deep Investigation

When fixes fail repeatedly, the deep investigation agent activates:

User: "I've tried 3 solutions but still getting this Next.js error"
Claude: [Performs exhaustive research across official docs, GitHub issues, Stack Overflow]

Fake Implementation Detection

User: "Check this codebase for fake implementations"
Claude: [Scans for hollow code patterns, placeholder implementations]

Core Principles

1. Single Responsibility & Separation of Concerns

• One function, one purpose
• One hook per concern
• Components render UI; business logic belongs in hooks or services

2. Pragmatic Library Adoption

• Search for existing solutions before implementing generic functionality
• Evaluate: community adoption, maintenance, TypeScript support, bundle size

3. Modular Architecture

• Feature-based organization over technical layering
• High cohesion, low coupling
• Export through barrel files to define public API

4. Type-First Development

• Types as specifications
• Discriminated unions over optional properties
• Branded types for domain identifiers

5. Functional Programming Style

• Pure functions, immutability
• Function composition
• Functions as primary unit of abstraction

Requirements

• Claude Code CLI
• For deep-investigation agent: Context7 MCP server (optional, for versioned documentation)
• For codex-reviewer agent: Codex MCP server (optional)

License

MIT