← ClaudeAtlas

android-harnesslisted

Use when working on this Android app architecture, Compose or XML UI, Room, performance, or daily checkout facts.
rabee-elkholy/android-agent-harness · ★ 6 · Testing & QA · score 81
Install: claude install-skill rabee-elkholy/android-agent-harness
# Android harness (domain knowledge) Deterministic architecture facts are derived during setup into `.agents/project-context/`. ## Authoritative Truth Resolution Hierarchy 1. **Current source evidence**, when inspected directly in checkout files. 2. **Generated project facts**: `.agents/project-context/project-facts.json` and rendered views (`architecture.md`, `ui.md`, `persistence.md`, `conventions.md`). 3. **Developer notes**: `.agents/project-context/project-notes.md` (domain intent & unresolved conventions). Developer notes clarify ambiguities or override generic assumptions, but may NOT contradict deterministic source facts. 4. **Generic Android guidelines**: Kit-owned reference rules below. ## Harness Execution Boundary During normal Android application tasks, installed harness engine source under `.agents/scripts/**` is an implementation detail. Do not inspect or recursively read harness Python implementation before executing a documented harness command. Use the documented public command contract and its output directly. Harness source inspection is allowed only when: 1. a documented harness command fails unexpectedly; 2. its output violates the documented contract; 3. Doctor reports harness corruption/inconsistency; 4. the developer explicitly asks to inspect, debug, or modify the harness itself. Application source inspection remains unaffected. For documented commands, see [Public Command Contract](./references/command-contract.md). ## Task Guidance & Routing