test-app
Skill0 starsChanged 6 months agoMITView on GitHub
Automatically open a web app and run a QA test pass on it using the agent-browser CLI. Use whenever the user wants to test, QA, smoke-test, or check a web application or website they are building — including localhost dev servers and live/staging URLs. Triggers include "testa la mia app", "test my app", "open my app and check it", "does my site work", "QA this page", "controlla il sito", "verifica il login", "run a test pass", "check every page", "look for bugs", "take screenshots and check for errors". Drives the locally installed `agent-browser` program (works on Windows, macOS, Linux). Prefer this over vision-based clicking when checking a live URL.
- Reads credentials
- Installs packages
4 other projects use a version of this.
Adopted by
Most recently taken up by
- Pdiamondz1/hma_project_foundationtheir own version, first seen Oct 9, 2026
- astashov/liftosaurtheir own version, first seen Oct 8, 2026
- MarkTripoli/skillstheir own version, first seen Sep 25, 2026
- ETdoFresh/ClaudeCodeOrchestratortheir own version, first seen Jan 10, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- astashov/liftosaur★ 775same name
- MarkTripoli/skills★ 2same name
- ETdoFresh/ClaudeCodeOrchestrator★ 1same name
- Pdiamondz1/hma_project_foundation★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 4 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
If agent-browser --version errors, stop and install it before testing.
- Distinctive: line 27
When the user just says "testa la mia app" (test my app) without a URL, figure out where it runs on your own.
- Distinctive: line 61
Use the refs from the snapshot to interact deterministically:
- Distinctive: line 78
Always Read the screenshots — visually confirm layout, empty states, overlaps, broken images.
- Distinctive: line 86
Never ask the user to paste passwords into chat.
- Very specific: line 105
Use --json when you want machine-readable output to parse.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
test-app — automated QA pass with agent-browser
Open a web app in a real browser, click through it, screenshot key pages,
check the console for errors, and report what works and what is broken.
This skill is the method; the engine is the agent-browser CLI.
Requirements (one-time)
The agent-browser program must be installed:
npm install -g agent-browser
agent-browser install # downloads the test Chrome, first time only
agent-browser --version # should print a version, e.g. 0.27.0If agent-browser --version errors, stop and install it before testing.
1. Find the app automatically — do NOT ask the user for the URL first
When the user just says "testa la mia app" (test my app) without a URL, figure out where it runs on your own. Only ask the user as a last resort.
a) If the user gave a URL or port, use it directly.
b) Otherwise, look for clues in the project files (current working dir):
package.jsonscripts / framework config → infer the dev port (Vite → 5173, Next.js → 3000, Astro → 4321, CRA → 3000, Vue CLI → 8080, Angular → 4200, SvelteKit → 5173)..env/ config files that set a PORT.
c) Probe the common local dev ports and use the first that responds:
for p in 3000 5173 8080 4321 4200 5000 8000 3001 4000; do
curl -s -o /dev/null -w "%{http_code}" "http://localhost:$p" --max-time 1 \
&& echo " <- localhost:$p is up"
donePick the first port that answers (any HTTP status counts as "running").
d) Only if nothing is found, tell the user the app does not seem to be running and ask them to start the dev server (or give the URL/port).
Confirm the chosen URL in one short line before testing, e.g. "Trovata l'app su http://localhost:5173 — la testo."
2. Open the page and read its structure
agent-browser open <url>
agent-browser wait --load networkidle
agent-browser snapshot -i # interactive elements with @e1, @e2 refs3. Walk the key pages and flows
Use the refs from the snapshot to interact deterministically:
agent-browser click @e2 # click by ref
agent-browser fill @e3 "test@example.com" # fill an input
agent-browser find text "Sign In" click # or locate by visible textRe-snapshot after any action that changes the page. Cover at minimum: the homepage, the main navigation targets, and any flow the user mentioned (signup, login, checkout, etc.). On a code branch, prefer the pages whose files changed.
4. Capture evidence on each page
agent-browser screenshot /tmp/page.png # then Read the image to inspect it
agent-browser console # JS console: log / warn / error
agent-browser errors # uncaught JavaScript exceptionsAlways Read the screenshots — visually confirm layout, empty states, overlaps, broken images. Console errors are bugs even when the page looks fine.
5. Pages behind a login
- Auth header for the API origin:
agent-browser open <url> --headers '{"Authorization":"Bearer <token>"}' - Or reuse an existing Chrome login profile:
agent-browser --profile Default open <url>Never ask the user to paste passwords into chat.
6. Report
Short QA report:
- Health: pages tested, how many passed.
- Issues by severity (Critical / High / Medium / Polish), each with the page, what is wrong, and the evidence (screenshot / console line).
- Flows verified end to end (e.g. signup → dashboard). Concrete, no flattery — surface real breakage.
7. Clean up
agent-browser closeEfficiency tips
- The browser persists between commands (background daemon), so chain with
&&when intermediate output is not needed:agent-browser open <url> && agent-browser wait --load networkidle && agent-browser snapshot -i - Use
--jsonwhen you want machine-readable output to parse. - Re-snapshot after navigation; stale refs point at the old page.
Severity guide
- Critical: data loss, broken core flow (cannot sign up / pay / submit), crash.
- High: feature visibly broken, console errors on load, broken auth.
- Medium: layout breaks at some widths, minor flow friction.
- Polish: spacing, copy, small visual inconsistencies.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

