From agentiqa
Test web and mobile apps using Agentiqa. Use when the user asks to test, QA, explore, or find bugs in their app. IMPORTANT: Do NOT explore the codebase or read source files — just run the global `agentiqa` CLI directly. Steps: (1) check CLI installed with `command -v agentiqa`, (2) install Chromium with `npx playwright install chromium`, (3) run `agentiqa explore "<prompt>" --url <url> --auto-approve --json` in background. NEVER use local repo builds like `node apps/cli/dist/cli.js`.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agentiqa:agentiqa-testThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You have access to `agentiqa explore` — a CLI that launches an independent AI testing agent to find bugs in web and mobile apps.
You have access to agentiqa explore — a CLI that launches an independent AI testing agent to find bugs in web and mobile apps.
ls, find, rg, cat, or read any source files. The agentiqa CLI is an external tool, not part of this repo. Go straight to Step 0.agentiqa CLI — run agentiqa explore ... directly. NEVER run node apps/cli/dist/cli.js, npx agentiqa, or any local build from the repo. The globally installed CLI has its own bundled Chromium and dependencies.The testing agent:
YOU MUST FOLLOW THESE STEPS IN ORDER. DO NOT SKIP ANY STEP.
Step 0: Ensure CLI is installed and authenticated (MANDATORY — run in foreground)
Run these checks EVERY TIME before launching explore. Do NOT skip.
# Install CLI if missing (MUST be global install)
command -v agentiqa >/dev/null 2>&1 || npm install -g agentiqa
# Check auth — if not logged in, run login (opens browser)
agentiqa whoami >/dev/null 2>&1 || agentiqa login
# Check ffmpeg for video recording (optional — screenshots work without it)
if ! command -v ffmpeg >/dev/null 2>&1; then
echo "[agentiqa] ffmpeg not found — video recording will be disabled. Install with: brew install ffmpeg"
fi
YOU MUST RUN BOTH CHECKS AND WAIT FOR THEM TO COMPLETE. If agentiqa login opens a browser, wait for the user to complete authentication before proceeding.
Step 1: Install Chromium (MANDATORY before first explore — run in foreground, wait for it to finish)
npx playwright install chromium
YOU MUST RUN THIS AND WAIT FOR IT TO COMPLETE BEFORE STEP 2. If you skip this, the explore command wastes 30+ seconds installing Chromium mid-run. On subsequent tests in the same session you can skip this step.
If the install times out or fails due to network restrictions, you may need to approve network access and retry.
Step 2: Launch explore in BACKGROUND (NEVER foreground — it takes 3-10 minutes and will time out)
agentiqa explore "<prompt>" --url <url> --auto-approve --json
Run this command in the background. Note the background command output file path.
Step 3: Monitor until completion
CRITICAL SUPER IMPORTANT: Monitoring background commands:
while ! grep -q "\[agentiqa\] Done" {backgroundCommandOutputFilePath} 2>/dev/null && ! grep -q "^Exited with code:" {backgroundCommandOutputFilePath} 2>/dev/null; do sleep 2; done
This polls until the CLI prints its [agentiqa] Done line OR the process exits. If the foreground monitor times out, just re-run the same while command — it will resume waiting.
Step 3: Read results
Once the while-loop exits, use tail to read the last lines of the background output file. The JSON result will be at the end of stdout.
IMPORTANT:
Do NOT use 2>&1 — keep stderr separate so the JSON on stdout stays clean.
"Test the checkout flow"If the user's message contains a URL (e.g. https://..., http://..., localhost:...), you MUST extract it and pass it as --url <extracted-url>.
Do NOT put the URL inside the prompt string — it must be a separate --url flag.
When no --target is specified, the CLI auto-detects running Android emulators and iOS simulators.
You do NOT need to specify --target, --package, or --device — just provide the prompt.
| Flag | Use when... |
|---|---|
--url <url> | Testing a web app — always pass when user provides a URL |
--target android|ios|web | Forcing a specific platform |
--package <name> | Launching a specific Android app before testing |
--bundle-id <id> | Launching a specific iOS app before testing |
--feature "<text>" | Describing what was built (user perspective) |
--hint "<text>" | Suggesting specific things to test (repeatable) |
--known-issue "<text>" | Suppressing known issues (repeatable) |
--credential "name:secret" | Providing login credentials (repeatable) |
--realtime | Using realtime agent for mobile (default: classic) |
--no-orchestrator | Bypassing orchestrator in realtime mode |
--no-artifacts | Skipping screenshot/video saving to temp dir |
--verbose | Showing detailed action/observation logs |
Test a web app (ALWAYS use background mode):
# Step 1: Launch in background
agentiqa explore "Test the signup flow" --url http://localhost:3001/signup --auto-approve --json
# Step 2: Monitor (replace {backgroundCommandOutputFilePath} with actual path)
while ! tail -n 2 {backgroundCommandOutputFilePath} | grep -q "^Exited with code:"; do sleep 1; done
# Step 3: Read results from the output file
Test with context:
agentiqa explore "Test the new checkout" \
--url http://localhost:3001/checkout \
--auto-approve --json \
--feature "Shopping cart checkout with Stripe" \
--hint "Try invalid card numbers" \
--hint "Try empty cart" \
--known-issue "Payment is in test mode"
stdout contains JSON:
{
"summary": "Tested the Settings screen. Found 2 issues...",
"issues": [
{
"title": "Save button stays disabled",
"severity": "high",
"category": "logical",
"confidence": 0.9,
"steps": ["Toggle switch", "Tap Save", "Nothing happens"]
}
],
"actionsTaken": 12,
"durationSeconds": 68,
"target": "android",
"device": "emulator-5554",
"artifactsDir": "/tmp/agentiqa-abc123",
"screenshotCount": 8
}
Artifacts are saved by default:
screenshot-001.png, screenshot-002.png, etc.recording.mp4 (output includes videoPath instead of screenshotCount)--no-artifacts to disablestderr contains progress lines (visible in the background task output).
After the command completes, read the output and present to the user:
If no issues found, say so — the app looks good!
If the command fails:
agentiqa login (opens browser for authentication). Do NOT tell the user to do it manually — run it yourself and wait for completion.npx playwright install chromium yourself — do not ask the user to do it manually.Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub agentiqa/agentiqa-plugin --plugin agentiqa