App Store Product Researcher

An Agent Skill that validates App Store niche ideas using real competitor data, user reviews, and revenue estimates — then delivers a go/no-go verdict.

Compatible with 30+ AI coding agents including Claude Code, OpenAI Codex, Cursor, and Gemini CLI.

Pipeline

Niche → Competitors → Reviews → Gap → App Idea → Validation Score (1-10)

1. Competitors — Multi-keyword iTunes search + Sensor Tower revenue/downloads. Direct vs. Adjacent tagging.
2. Reviews — 50 recent reviews per app. Recent sentiment vs. overall rating to detect declining apps.
3. Gap — Biggest unmet need, consistent failures, switching triggers.
4. Idea — Name options, features mapped to complaints, onboarding, monetization, viral concept.
5. Score — Pain / Market / Gap / Monetization / Build difficulty. Honest build-or-skip verdict.

Quick Start

# Claude Code
/app-store-product-researcher "Quit Vaping"

# Any agent — place repo in project, then prompt:
# "Research the 'Quit Vaping' niche using the scripts in scripts/"

Scripts

Free APIs, no auth. Work standalone or via agents.

bash scripts/search_apps.sh "habit tracker" [country] [limit]   # iTunes Search API
bash scripts/fetch_app_stats.sh "ID1,ID2,ID3"                   # Sensor Tower estimates
bash scripts/fetch_reviews.sh APP_ID [country] [page]            # iTunes RSS reviews

Example

From a "Digital Detox" run:

App	Downloads	Revenue	Overall	Recent	Signal
Opal	300k	$700k	4.79	3.42	Declining
BePresent	200k	$300k	4.84	3.02	Severe decline
ScreenZen	100k	$30k	4.85	4.88	Thriving

Score: 7.0/10 — "Build if you have a differentiation story. ScreenZen owns the honest-free position."

Multi-Agent

Supports parallel execution (~2x faster, tested):

Coordinator
  ├── 3 Search Agents (parallel)   → deduplicate → Sensor Tower batch
  ├── 3 Review Agents (parallel)   → structured summaries
  └── Synthesize → Gap → Idea → Score

Implementation

Define subagents in .claude/agents/. Each gets clean context + access to scripts/ via ${CLAUDE_SKILL_DIR}. Subagents can't nest.

Scenario	Approach
Single niche	Sequential or parallel data collection
Comparing niches	One agent per niche, coordinator ranks
Quick app check	Standalone scripts

Design

Tested across 6 niches (Quit Vaping, Anxiety Relief, Fasting, White Noise, Digital Detox, Pet Training):

• 3+ search variations — single keywords miss 30-50% of competitors
• Direct vs. Adjacent — adjacent giants inflate market size 5-20x
• Recent sentiment check — most valuable signal; every niche had declining leaders
• Bundled scripts — deterministic API work, not reinvented per run

Limitations

• Revenue/downloads are Sensor Tower estimates (ML models, coarse buckets)
• Reviews cover the most recent 50 — days for popular apps, months for small ones
• US App Store default; pass country code for other markets
• iOS only; Google Play needs different sources

License

MIT