claude-elixir-phoenix: Instructions file for Claude Code

Plugin Development Guide

Development documentation for the Elixir/Phoenix Claude Code plugin.

Overview

This plugin provides agentic workflow orchestration with specialist agents and reference skills for Elixir/Phoenix/LiveView development.

Workflow Architecture

The plugin implements a Plan → Work → Review → Compound lifecycle:

/phx:plan → /phx:work → /phx:review → /phx:compound
     │           │            │              │
     ↓           ↓            ↓              ↓
plans/{slug}/  (in namespace) (in namespace) solutions/

Migration note: The --depth flag replaces the old --detail flag. Use quick|standard|deep instead of minimal|more|comprehensive.

Key principle: Filesystem is the state machine. Each phase reads from previous phase's output. Solutions feed back into future cycles.

Workflow Commands

Command	Phase	Input	Output
/phx:plan	Planning	Feature description	plans/{slug}/plan.md
/phx:plan --existing	Enhancement	Plan file	Enhanced plan with research
/phx:brief	Understanding	Plan file	Interactive walkthrough (ephemeral)
/phx:work	Execution	Plan file	Updated checkboxes, plans/{slug}/progress.md
/phx:review	Quality	Changed files	plans/{slug}/reviews/
/phx:compound	Knowledge	Solved problem	solutions/{category}/{fix}.md
/phx:full	All	Feature description	Complete cycle with compounding

Artifact Directories

Each plan owns all its artifacts in a namespace directory:

.claude/
├── plans/{slug}/              # Everything for ONE plan
│   ├── plan.md                # The plan itself
│   ├── interview.md           # Brainstorm → plan contract (requirements handoff)
│   ├── research/              # Research agent output
│   ├── reviews/               # Review agent output (individual tracks)
│   ├── summaries/             # Context-supervisor compressed output
│   ├── progress.md            # Progress log
│   └── scratchpad.md          # Auto-written decisions, dead-ends, handoffs
├── audit/                     # Audit namespace (not plan-specific)
│   ├── reports/               # 5 specialist agent outputs
│   └── summaries/             # Supervisor compressed output
├── reviews/                   # Fallback for ad-hoc reviews (no plan)
├── skill-metrics/             # Skill effectiveness dashboards and recommendations
│   ├── dashboard-{date}.json  # Per-skill aggregate metrics
│   └── recommendations-{date}.md  # Improvement recommendations
└── solutions/{category}/      # Global compound knowledge (unchanged)
    ├── ecto-issues/
    ├── liveview-issues/
    └── ...