Pose Match

Mobile AR app that takes a reference photo (Pinterest pose, etc.) and overlays it onto a live camera feed, guiding subject and photographer into the matching shot.

Project docs

• architecture.md — full system architecture
• tasks.md — sequential MVP build plan, one atomic task at a time

Working with Cursor

This repo has .cursor/rules/ configured so any Cursor agent opening it automatically loads:

• project-context.mdc — project overview, tech stack, non-negotiables
• workflow.mdc — the one-task-at-a-time discipline tied to tasks.md
• coding-style.mdc — TS conventions, state tier discipline, naming

You don't need to share a chat between collaborators. Both of you get the same agent behavior because both Cursor instances read the same rules from the repo.

Starting work in Cursor

Open a new chat. Say one of:

• do the next task in tasks.md — agent picks up from the next unchecked task
• do task 7.3 — agent does that specific task
• what's the next task? — agent reports without starting

The agent will state the task, execute it, run the test, and stop. You verify, then ask for the next one.

Collaboration model

• Both collaborators have access to the repo.
• Claim tasks in tasks.md by tagging your name (e.g., ### 7.3 [@om]).
• Commit after each task so the other person's agent picks up the latest state.
• For hand-offs of tricky work, share a Cursor transcript link in chat — the other person starts a fresh agent chat referencing it.

Repo structure (target after build)

pose-match/
├── apps/mobile/             # React Native app (Expo bare workflow)
├── services/pose-worker/    # Python keypoint extraction service
├── packages/
│   ├── pose-math/           # Pure TS comparison logic
│   └── shared-types/        # Shared TS types
└── supabase/
    ├── migrations/          # Postgres schema + RLS
    └── functions/           # Edge functions (Deno/TS)

See architecture.md for the full breakdown.

Status

Pre-MVP. Building per tasks.md, starting at Phase 0.