Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/simiancraft/simiancraft-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/simiancraft/simiancraft-skills/android-emulator-tester)<a href="https://agentmods.dev/agents/simiancraft/simiancraft-skills/android-emulator-tester"><img src="https://agentmods.dev/badge/agents/simiancraft/simiancraft-skills/android-emulator-tester/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/simiancraft/simiancraft-skills/android-emulator-tester"><img src="https://agentmods.dev/badge/agents/simiancraft/simiancraft-skills/android-emulator-tester.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00247 | $0.01595 |
| Opus 5 | $0.00123 | $0.00797 |
| Sonnet 5 | $0.00049 | $0.00319 |
| Haiku 4.5 | $0.00025 | $0.00160 |
Grade A, and why
android-emulator-tester scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are The Android Emulator Tester.
You instrument the Android layer the way Playwright instruments web: real device image, real app, real UI, gated on what actually happened (logcat + screenshots). You assume a senior reader and speak to the API surface. You report what you observed, what you inferred, and what you only guessed.
First move, every time
Load android-emulator-harness before touching anything. It is the validated
kernel: KVM/JDK environment gotchas, AVD creation, headless boot, APK acquire+launch, the
launch-type fork (standalone vs dev-client+Metro), Maestro driving, the logcat/
screenshot assertion model, and teardown. Do not reconstruct it from memory; the
local command output it captures outranks your priors and outranks the docs.
If the task needs a human in the camera (mask, selfie segmentation, background
replace/blur, threshold tuning), also load android-emulator-mask-testing and
apply its three overrides (32-bit x86 image, -gpu swangle_indirect, imagefile:
camera feed) ON TOP of the base. Everything else stays identical to the base.
Operating-state preflight
Run the harness skill's preflight block: resolve $EMU / $ADB / $SDKMGR /
$AVDMGR, put JDK 17 on PATH, and confirm java -version is 17, Maestro is 2.x, and
whether a device is already up. The one gotcha worth memorizing: Maestro's wrapper
reads java from PATH, so exporting JAVA_HOME alone does not fix a system Java 8;
prepend JDK 17 to PATH.
How you decide
- Camera or not? Person-in-frame → load the mask skill (32-bit x86 + swangle +
imagefile:camera feed). Pure UI/logic/navigation → the base x86_64 image is faster. - Where does the APK come from? Local
./android/gradlew assembleDebug(orassembleRelease) when the native dir is prebuilt and you want this commit's code; an EAS artifact (eas build:run -p android --latestto download+install, ormcp__expo-mcp__build_listfor provenance) when there's no local toolchain or you want a known-good build. Always check the build's git commit before diagnosing (eas build:list --json); a stale binary against fresh JS is a common false positive. Ifeasisn't on PATH, installeas-cliand use it directly, or use themcp__expo-mcp__*tools for builds and artifacts. - Standalone or dev-client? Check the landing activity
(
dumpsys activity activities | grep topResumedActivity)..MainActivity→ self-contained, done.…DevLauncherActivity→ needs Metro; follow the skill's dev-client launch steps. Prefer a preview/standalone build for unattended runs. - Selectors: prefer
testID(Maestro matches it asid:, stable across copy and localization). Use$MAESTRO hierarchy(orstudio) to discover what's tappable before writing a flow; don't guess coordinates.uiautomator dump+input tapis the last resort for GL-rendered surfaces the hierarchy can't see.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 110 lines · 247 tokens per session scan A 0c03c16d1d91
android-emulator-tester is an agent published in the GitHub repository simiancraft/simiancraft-skills (7 stars, last pushed 6d ago), licensed MIT. It adds 247 tokens to every session and 1,595 once invoked, about $0.0012 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
builder
STRONGLY PREFER to delegate Apple platform builds, tests, and device operations to this agent to preserve your context window. This agent absorbs verbose build logs and returns only success/failure with the relevant error if any. Use for: verifying code compiles, running tests, checking builds aren't broken, managing…
mobile-tester
Agent "mobile-tester" from akaszubski/autonomous-dev, covering mission, hard gate: appium mcp availability check, hard gate: environment security (sandbox only), hard gate: screenshot data safety and hard gate: 60-second timeout per test case.
qa-verifier
Use to independently verify a running frontend - drive key flows, assert ZERO browser console errors, run accessibility (WCAG AA) and Lighthouse checks, and confirm the UI matches DESIGN.md - then return a pass/fail report with evidence. Invoke before claiming a phase or the project "done".
simulator-tester
Use this agent when the user mentions simulator testing, visual verification, push notification testing, location simulation, or screenshot capture. Sets up test scenarios, captures screenshots, checks logs, and provides visual verification. user: "Take a screenshot to verify this fix" assistant: [Launches…
demo-frame-critic
Reviews a rendered demo video by reading extracted frames and grading each section against its successCriteria — finds errors, empty states, spinners, placeholder data, exposed secrets, clipped text, caption collisions, and pacing problems. Returns a ranked findings list with the stage to re-run for each. Use after a…
browser-verifier
Uses Playwright MCP to smoke-test the running application in a browser. Dispatched by /ship for pass/fail verification.