nicholasrae-review-reply
di nicholasrae
Marketing & SalesSintesi
Automatically monitors your App Store reviews and drafts warm, on-brand replies for 1–3 star reviews — so unhappy.
Descrizione
Cosa fa questa skill
Monitors App Store Connect for new reviews and drafts replies for 1-3 star ratings using Claude. Repeat complaints appearing 3+ times in 7 days trigger immediate bug alerts. A daily 8am Telegram digest lets you approve, edit, or reject each draft before it posts.
It automates the monitoring and drafting while keeping a human approval step before anything posts, avoiding the risk of a bot reply going live unreviewed.
Quando usarla
- Catching a wave of crash complaints before your rating drops
- Approving reply drafts from your phone each morning
- Spotting a broken login flow from clustered 1-star reviews
- Maintaining response rate across several iOS apps at once
- Drafting a warm reply to a frustrated user within hours of their review
Esempio di workflow
User asks: check for new reviews
- 1
Runs monitor.py, which calls the App Store Connect API for each configured app
- 2
Stores new reviews in data/reviews.json, skipping IDs already seen
- 3
Triggers draft_reply.py for every review rated 1-3 stars
- 4
Passes all new reviews to pattern_detector.py to identify recurring complaint themes
- 5
Adds draft replies to data/queue.json with status pending
New reviews fetched and logged; drafts queued for tomorrow's 8am Telegram digest; any repeat-complaint pattern above threshold triggers an immediate alert
Requisiti
- App Store Connect account with API access enabled
- APP_STORE_KEY_ID environment variable
- APP_STORE_ISSUER_ID environment variable
- APP_STORE_PRIVATE_KEY_PATH environment variable (path to .p8 private key downloaded from App Store Connect)
- ANTHROPIC_API_KEY environment variable
- Telegram bot token (create via BotFather)
- TELEGRAM_BOT_TOKEN environment variable
- TELEGRAM_CHAT_ID environment variable
Installazione
clawhub install nicholasrae/nicholasrae-review-replyEsegui il comando CLI oppure incolla il prompt in Claude Code, Cursor, Codex o nell'agente di coding che usi.