aiatsuk

Autonom — Testing skill for Claude Code

Testing community

Universal mobile test and debug harness for AI coding agents on Android emulators and iOS Simulator, with semantic UI automation, repeatable flows, diagnostics, and evidence-rich reports.

How to install Autonom

This entry records only its repository, not the path inside it, so there is no exact command to give. Open aiatsuk/autonom and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Autonom does

Universal mobile test and debug harness for AI coding agents on Android emulators and iOS Simulator, with semantic UI automation, repeatable flows, diagnostics, and evidence-rich reports.

Alternatives in Testing

Debug — You are tasked with helping debug issues during manual testing or implementation 10k ★
Skill Debug — Debug issues methodically — use when stuck on errors, test failures, or unexpected behavior 2.8k ★
Claude Codepro — by Max Ritter - Professional development environment for Claude Code with spec-driven workflow, TDD enforcemen 1.6k ★

README

Autonom

**Universal mobile test and debug harness for AI coding agents.**

Autonom gives Codex, Claude, Grok, and other skill-compatible agents repeatable workflows to **test and debug Android and iOS apps**: own a target session, see what is on screen (compact accessibility tree + screenshot), tap/type/gesture by semantics, pull logs and crash reports, drive device state, and climb an evidence ladder through builds, UI checks, performance, memory, and release validation.

The same verbs work on an Android emulator and an iOS Simulator, and the compact node schema is identical on both, so one skill body drives either platform.

Skills are portable `SKILL.md` packages. The **CLI control plane** — installed as `autonom` — is the stable JSON API. Helpers stay dependency-light. No MCP server is required (optional MCP wrapper is planned).

**Shipped:** Android emulator + iOS Simulator sessions, UI trees, semantic find/tap/gestures, screenshots and recordings, logs and crashes, deep links, permissions, location, media and container files, consent-gated HTTP(S) capture and mocking, a per-session journal of every action, and **Flow v1** — repeatable flow files with exact selectors, polling assertions, and failure classes ([`docs/FLOW.md`](docs/FLOW.md)), plus Maestro import/export, Teach recording and approved App Skills, addressable per-step evidence reports (HTML/JUnit/Allure/agent/CSV/metrics), portable integrity-checked Report Bundle v2, baseline replay/checkpoints, supervised CI campaigns, Mobile Canvas, an observed Runtime Map, local PR proof (`autonom proof --base`), live session watch (`session outputs`, `logs follow`, `network requests follow`), and an `autonom metrics` family (memory, CPU, frames, traces), and deterministic capture state — a pinned status bar on both platforms, a pinned keyboard/locale on iOS, and system animations off on Android — with a `repair` hand-off (including ranked on-screen candidates) on every failed flow step. iOS input runs through