elixir-expert

SKILL.md

Elixir Expert

Purpose

Provides expertise in Elixir development, Phoenix Framework, and OTP patterns. Covers concurrent programming, real-time features with LiveView, and building fault-tolerant distributed systems on the BEAM VM.

When to Use

Building Elixir applications

Developing Phoenix web applications

Implementing real-time features with LiveView

Using OTP patterns (GenServer, Supervisor)

Building distributed systems on BEAM

Designing fault-tolerant architectures

Working with Ecto for database access

Quick Start

Invoke this skill when:

Building Elixir applications

Developing Phoenix web applications

Implementing real-time features with LiveView

Using OTP patterns

Designing fault-tolerant systems

Do NOT invoke when:

Building Ruby on Rails apps (use rails-expert)

Building Node.js backends (use javascript-pro)

Building Python backends (use python-pro)

Infrastructure automation (use terraform-engineer)

Decision Framework

Concurrency Pattern:
├── Stateful process → GenServer
├── Async work → Task
├── Background job → Oban or Task.Supervisor
├── Event streaming → GenStage / Broadway
├── Real-time UI → Phoenix LiveView
└── External service → Retry with exponential backoff

Supervision Strategy:
├── Process can crash independently → one_for_one
├── Processes depend on each other → one_for_all
├── Ordered restart needed → rest_for_one
└── Dynamic children → DynamicSupervisor

Core Workflows

1. Phoenix Application Setup

Generate Phoenix project

Configure database with Ecto

Define schemas and migrations

Create contexts for business logic

Build controllers or LiveViews

Add authentication

Deploy with releases

2. OTP Application Design

Identify stateful components

Design supervision tree

Implement GenServers for state

Add proper error handling

Implement graceful shutdown

Test supervision strategies

3. Real-Time with LiveView

Generate LiveView module

Define assigns and state

Implement handle_event callbacks

Use pubsub for broadcasts

Optimize with temporary_assigns

Add JS hooks if needed

Best Practices

Let it crash - design for failure recovery

Use supervision trees for fault tolerance

Keep GenServer state minimal

Use contexts to organize business logic

Prefer immutable data transformations

Test concurrent code with async: true

Anti-Patterns

Anti-Pattern
Problem
Correct Approach

Large GenServer state
Memory and serialization
External storage, ETS

Defensive coding
Hides bugs
Let it crash, supervise

Blocking GenServer
Process bottleneck
Async tasks for I/O

No supervision
Unrecoverable crashes
Proper supervision tree

Mutable mindset
Bugs and race conditions
Embrace immutability