Mobile Starter Kit
Agent skills

Agent skills

The task-specific playbooks bundled inside the downloaded mobile starter.

Agent skills are markdown playbooks in agent-skills/. They are written for your AI coding tool, not as commands you must learn. Ask for an outcome in plain language; AGENTS.md tells the agent which skill to read.

Start and customize

Ask forSkill the agent reads
“Get this running on my phone”get-it-running.md
“Make this app mine”make-it-yours.md
“Show me what needs a real provider”integration-points.md
“Set up GitHub backup”setup-github-backup.md
“Replace the icon and splash”app-icon-and-splash.md
“Add another language”add-i18n.md
“Add a date picker/sheet/component”add-ui-components.md

Activate providers

Ask forSkill the agent reads
“Activate all production integrations safely”setup-production-integrations.md
“Connect Supabase auth and data”wire-auth.md
“Activate account deletion”wire-account-deletion.md
“Connect RevenueCat”wire-payments.md
“Connect Coach and Scan AI”wire-ai-chat.md
“Activate voice transcription”wire-ai-voice.md
“Use private Supabase uploads”wire-file-uploads.md
“Activate Expo Push”wire-push.md
“Activate PostHog”wire-analytics.md
“Activate Sentry”wire-error-tracking.md

Native product work

Ask forSkill the agent reads
“Change Expo/native behavior”expo-best-practices.md
“Set up deep/universal links”add-deep-linking.md
“Instrument onboarding”add-onboarding-analytics.md
“Add an in-app review prompt”add-review-prompt.md
“Add OTA updates”add-ota-updates.md

Audit and ship

Ask forSkill the agent reads
“Review this screen against the design”project-design-review.md
“Audit accessibility”a11y-audit.md
“Audit performance”perf-audit.md
“Audit security”security-review.md
“Test the real app on devices”test-on-device.md
“Set up EAS”setup-eas.md
“Create store screenshots and copy”store-assets.md
“Run the final release gate”pre-launch-checklist.md
“Submit to Apple”submit-ios.md
“Submit to Google Play”submit-android.md

The skill workflow

Your request → AGENTS.md chooses a skill → agent reads it completely →
agent inspects ready code → you approve external/account actions →
agent implements → agent runs the skill's verification → you review evidence

Skills do not grant permission to deploy, buy an account, submit an app, or change a team-owned cloud project. They preserve those approval boundaries.

On this page