How do I install the App Test skill?

Run npx skillmds@latest add zedarvates/app-test in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is App Test safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with App Test?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is App Test free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published App Test?

zedarvates (@zedarvates) published this skill. Their other Agent Skills are listed on their SkillMD profile.

App Test by zedarvates — Agent Skill

Name: App Test
Author: zedarvates

App Test

Local-first GUI/app testing by image matching (SikuliX) — turn a small JSON spec referencing your button images into a runnable SikuliX script and run it locally, using a vision NPU (Hailo-8/10) or local vision model instead of cloud vision. Use when the user wants to test a desktop/game/web app "for real" by clicking buttons, has button images already, or mentions SikuliX, image-matching tests, or local UI tests.

zedarvates fbb4234 6 files · 20.9 KB Updated Sep 17, 2026

File contents

app_test — test apps locally by clicking their buttons

When you build a UI you already have the button images, so an image-matching bot can drive it — no cloud vision tokens needed. A tiny JSON spec → a runnable SikuliX script.

Spec → script → run

python -m skills.app_test.cli gen tests/login_flow.json     # print the SikuliX script
python -m skills.app_test.cli run tests/login_flow.json --out build

{
  "name": "login_flow",
  "image_dir": "tests/images",
  "similarity": 0.8,
  "steps": [
    {"do": "wait",           "image": "login_btn.png", "timeout": 10},
    {"do": "click",          "image": "login_btn.png"},
    {"do": "type",           "text": "user@example.com"},
    {"do": "click",          "image": "submit.png"},
    {"do": "assert_visible", "image": "welcome.png", "timeout": 8},
    {"do": "assert_absent",  "image": "error.png"}
  ]
}

Actions: wait · click · double_click · right_click · type · sleep · assert_visible · assert_absent.

Running it

Uses OculiX (the maintained SikuliX fork: OpenCV matching + embedded Tesseract OCR, Java) — or any SikuliX. The generated -r <bundle> scripts are drop-in compatible with both.

Install OculiX (Java 11+ required): download the platform "ide" jar from https://github.com/oculix-org/Oculix/releases and place it at ~/.oculix/oculixide.jar (auto-detected), or set OCULIX_JAR=/path/to/jar. runsikulix/oculix on PATH also work. run always generates the .sikuli bundle; it executes it via java -jar <jar> -r <bundle> -c when a runner is found, else it reports how to install one.

Verified: a generated bundle runs end-to-end on OculiX 3.0.4 (oculixide-3.0.4-windows.jar) — Jython executes the steps and returns the exit code (0 = pass).

Why local / economical

Image matching is CPU-cheap and needs no model. For richer checks (is this screen semantically right?), point verification at a Hailo-8/10 NPU or a local vision model via [[llm_backends]] instead of a paid cloud vision API — 0 cloud tokens. The generator is deterministic and unit-tested; the GUI run is local.

Related: [[llm_backends]] (local vision backends), [[metrics]].

---name: app-testdescription: Local-first GUI/app testing by image matching (SikuliX) — turn a small JSON spec referencing your button images into a runnable SikuliX script and run it locally, using a vision NPU (Hailo-8/10) or local vision model instead of cloud vision. Use when the user wants to test a desktop/game/web app "for real" by clicking buttons, has button images already, or mentions SikuliX, image-matching tests, or local UI tests.---# app_test — test apps locally by clicking their buttonsWhen you build a UI you already have the button images, so an image-matching botcan drive it — no cloud vision tokens needed. A tiny JSON spec → a runnableSikuliX script.## Spec → script → run```bashpython -m skills.app_test.cli gen tests/login_flow.json # print the SikuliX scriptpython -m skills.app_test.cli run tests/login_flow.json --out build``````json{ "name": "login_flow", "image_dir": "tests/images", "similarity": 0.8, "steps": [ {"do": "wait", "image": "login_btn.png", "timeout": 10}, {"do": "click", "image": "login_btn.png"}, {"do": "type", "text": "user@example.com"}, {"do": "click", "image": "submit.png"}, {"do": "assert_visible", "image": "welcome.png", "timeout": 8}, {"do": "assert_absent", "image": "error.png"} ]}```Actions: `wait · click · double_click · right_click · type · sleep ·assert_visible · assert_absent`.## Running itUses **OculiX** (the maintained SikuliX fork: OpenCV matching + embeddedTesseract OCR, Java) — or any SikuliX. The generated `-r <bundle>` scripts aredrop-in compatible with both.Install OculiX (Java 11+ required): download the platform "ide" jar fromhttps://github.com/oculix-org/Oculix/releases and place it at`~/.oculix/oculixide.jar` (auto-detected), or set `OCULIX_JAR=/path/to/jar`.`runsikulix`/`oculix` on PATH also work. `run` always generates the `.sikuli`bundle; it executes it via `java -jar <jar> -r <bundle> -c` when a runner isfound, else it reports how to install one.Verified: a generated bundle runs end-to-end on OculiX 3.0.4(`oculixide-3.0.4-windows.jar`) — Jython executes the steps and returns the exitcode (0 = pass).## Why local / economicalImage matching is CPU-cheap and needs no model. For richer checks (is this screen*semantically* right?), point verification at a **Hailo-8/10 NPU** or a localvision model via [[llm_backends]] instead of a paid cloud vision API — 0 cloudtokens. The generator is deterministic and unit-tested; the GUI run is local.Related: [[llm_backends]] (local vision backends), [[metrics]].

Frequently asked questions