ex-appagent

EX-APPAgent

Autonomous app management agent for Claude Code. Analyzes, optimizes, and grows your apps toward $20/day revenue through data-driven iterative optimization.

Inspired by autoresearch — the loop never changes, the data evolves.

What It Does

• Competitor analysis via browser research and third-party platforms
• ASO keyword optimization suggestions
• Conversion funnel analysis
• Feature gap identification
• Revenue and pricing strategy
• Experiment tracking with keep/discard verdicts
• Cross-app experience accumulation

You set goals in program.md, the agent does the rest. You only approve plans and handle what AI can't do (App Store submissions, payments, etc.).

Install

claude plugins marketplace add Shameless521/EX-APPAgent
claude plugins install ex-appagent

Usage

Navigate to any app project directory and run:

/appagent                              # Auto-analyze, agent decides what to focus on
/appagent analyze competitor VSCO       # Specific request in natural language
/appagent how to improve conversion     # Strategy question
/appagent what did we do this week      # Progress report

On first run, the agent automatically:

1. Creates a program.md template for you to fill in
2. Initializes the .appagent/ working directory
3. Analyzes your code and competitors
4. Generates initial strategy recommendations

How It Works

The Harness Loop

Every /appagent run executes the same 10-step cycle:

1. Read program.md       → Your goals and constraints
2. Check health          → Is data fresh?
3. Read state            → Where are we?
4. Read metrics          → Latest numbers
5. Read experiments      → Any results?
6. Read experience       → What worked before?
7. Determine priority    → What matters most right now?
8. Execute               → Analyze and generate plans
9. Present summary       → Show you the results
10. Update state         → Record this cycle

Priority System

The agent automatically determines what's most important:

Priority	Trigger	Action
1	Milestone reached	Celebrate, unlock new strategies
2	Plans awaiting review	Present for approval
3	Experiment complete	Judge keep/discard
4	Data anomaly	Diagnose and respond
5	Normal	Push current focus forward

program.md

Your only job is defining the strategy in program.md:

# Identity
name: MyApp
positioning: AI photo editor for enthusiasts

# Target
north_star: Daily pure revenue $20
milestones:
  - $1/day → validated (unlock: ad spend)
  - $5/day → growing (unlock: higher budget)
  - $20/day → target achieved

# Competitors
watch_list:
  - VSCO (priority) — focus: pricing, new features
  - Snapseed — focus: free model monetization

# Guardrails
never_llm:
  - False advertising or fake reviews
  - Copy competitor UI designs

# Current Focus
priorities:
  1. ASO keyword optimization
  2. Improve trial-to-paid conversion

The agent reads this file but never modifies it. You change it when strategy shifts.

Architecture

Hybrid: Claude Code Skill + Python Engine (Phase 2)

┌─────────────────────┐     ┌─────────────────────┐
│  /appagent Skill     │     │  Python Engine       │
│  Decision Engine     │◄───►│  Data Collectors     │
│  Code Operations     │files│  Trend Analyzer      │
│  Plan Generation     │     │  Experiment Tracker  │
└─────────────────────┘     └─────────────────────┘
         ▲                           ▲
    You trigger                 cron/launchd
    /appagent                   (Phase 2)

Phase 1 (current): Skill layer only — competitor analysis via browser, code analysis, strategy generation, plan lifecycle.

Phase 2 (planned): Python engine for automated data collection via App Store Connect API.

Phase 3 (planned): Self-improvement system with cross-app experience migration and capability expansion.

Project Structure

plugins/ex-appagent/
├── commands/appagent.md            # Main skill entry point
├── modules/                        # Instruction modules
│   ├── harness-loop.md             # Core 10-step cycle
│   ├── cold-start.md               # First-run initialization
│   ├── priority-engine.md          # Priority determination
│   ├── action-lifecycle.md         # Plan approval workflow
│   └── state-manager.md            # State read/write
├── agents/                         # Sub-agent instructions
│   ├── analyst.md                  # Data analysis & strategy
│   ├── competitor-researcher.md    # Market research
│   └── code-operator.md           # Code modifications
└── templates/                      # Initialization templates
    ├── program-template.md         # program.md template
    ├── state-initial.json          # Initial state
    └── health-initial.json         # Initial health

Per-App Data