What makes the /phx:full skill different from standard code generation?

Unlike single prompts, this skill uses a structured state machine with 20 specialist agents to plan, implement, verify, and review features autonomously.

Is there a limit to how long the autonomous agent will run?

Yes, the skill includes configurable cycle limits such as --max-cycles and --max-retries to ensure the agent stops and reports status if it cannot resolve a blocker.

Can I use this for existing Phoenix projects?

Yes, it is designed to integrate with existing Elixir/Phoenix applications, assessing your current context and domain modules before planning new features.

What is the 'Compounding' phase mentioned in the workflow?

Compounding is an automated final step that captures solved problems and patterns into a searchable knowledge base in your repository for future use.

How does the skill handle code errors during implementation?

It enforces 'Iron Laws' which require every task to pass 'mix compile' and 'mix test' before proceeding, automatically looping back to fix issues found during verification.

Full Phoenix Feature Development

Name: Full Phoenix Feature Development
Author: oliver-kriska

byoliver-kriska

•

API開発

Automates the end-to-end implementation of complex Elixir and Phoenix features using a multi-agent autonomous workflow.

This skill provides a comprehensive, multi-phase pipeline for building large-scale features in Elixir, Phoenix, and LiveView. It coordinates twenty specialized agents to manage the entire software development lifecycle—from initial discovery and architectural planning to implementation, automated testing, and multi-agent peer reviews. By enforcing strict 'Iron Laws' of verification and capturing learnings into a searchable knowledge base, it ensures that new domain modules and complex contexts are production-ready, well-documented, and aligned with Elixir best practices.

主な機能

01Strict 'Iron Law' enforcement requiring code to pass all compiler and test checks

02Knowledge compounding that captures solved problems into a searchable solution base

03Parallel four-agent code reviews for specialized Elixir and Phoenix audits

0462 GitHub stars

05Automatic state tracking via progress.md and integrated Claude Code tasks

06Autonomous multi-agent planning and research for complex domain modules

ユースケース

01Refactoring or creating new domain modules with integrated background job processing

02Building real-time notification architectures using Phoenix PubSub and LiveView

03Implementing complex multi-context features like magic link authentication systems

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add oliver-kriska/claude-elixir-phoenix full

For use in Claude.ai and ChatGPT

Download Skill