appatlaslisted

# AppAtlas — map every screen of any app ## Overview You are the **orchestrator**: you resolve blockers, keep the index current, and dispatch Sonnet subagents to crawl and write. The product's UI surface (one Chrome tab group, one simulator) is a **single shared resource — crawl agents run strictly one at a time**; parallelism is for doc/review/fixer agents only. Read `references/playbook.md` in this skill directory before dispatching anything — it has the briefing template, capture workflows, recovery snippets, and index builder. **Iron rule of parallelism:** at most ONE crawl agent per physical UI resource at a time. The Chrome tab group is one resource (Chrome MCP acts on the focused tab; parallel agents interleave clicks and freeze screenshot capture — "one tab per agent, 4–6 in parallel" does not work). Each simulator/emulator is its own resource. Distinct resources (e.g. the web lane and an iOS-simulator lane) MAY run as concurrent lanes, each with its own briefing; within a lane, strictly sequential. ## Phases 1. **Recon (orchestrator, inline)** — sign in/up, complete onboarding *capturing every step as spec material* (fields, defaults, dropdown options, validation, side effects). Fill realistic demo data (a real brand/site you own beats lorem — features populate). Map the full nav → write the section list, folder tree `NN-section/{spec.md,screenshots/}`, `README.md` index with a status table + post-crawl checklist, and `AGENT-BRIEFING.md` from the playbook templa