Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Hostile Qa Tester

    1

    Use to QA a web app by acting as the most difficult, tech-resistant user: persona-driven sessions that ignore instructions, try unexpected paths, and break flow

    Free

    6 usesSecurity scanned
    Hostile Qa Tester

    Hostile Qa Tester

    Example session with this skill installed

    Run a hostile QA pass on our new checkout flow at https://staging.app/cart using 2 personas. Focus on interruptions and empty states.

    • Read your context and instructions
    • Compiled the hostile qa tester

    Session Complete. Found 3 bugs.

    • S1 Blocker: App crashes on double-click submit (01_bob_checkout.png).
    • S2 Major: Refreshing at Step 2 clears the entire cart (02_gina_refresh.png).
    • S3 Minor: Missing validation message for empty CVV.
      Full log and 8 screenshots saved to /qa-reports.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Hostile QA Tester

    When to Use

    Use this skill before a release, after a redesign, or when "it works for me" reports pile up. It role-plays the user your product team never invites: impatient, misclicking, tech-resistant, and determined to do the wrong thing. Output is a reproducible evidence log plus a severity-ranked bug report.

    Do not use for: accessibility conformance audits (run a dedicated a11y pass); performance and load testing; security penetration testing (this is UI behavior, not exploitation); products with no UI to attack, such as pure APIs.

    Procedure

    1. Pick personas. Choose at most 3, extreme but believable: "Boss Bob" (clicks the first button, reads nothing), "Grandma Gina" (wrong taps, huge fonts, no keyboard), "Hacker Harry" (types quotes and scripts into every field and expects graceful errors).

    2. Define scenarios from real user journeys. Per persona write 4–6 scenarios covering: the happy path with interruptions (back button, refresh mid-flow), empty states, extreme inputs (empty, 5000 chars, emoji, '; DROP TABLE), and cancellation paths (close a modal, sign out mid-checkout).

    3. Execute with hostility rules: never follow an error message literally; retry the failed action twice; click adjacent elements; submit forms without filling them; use browser back/forward aggressively; resize the window mid-interaction; ignore onboarding hints on purpose. Record every step.

    4. Capture evidence: screenshot before and after each suspicious action; note browser, OS, viewport, and the exact click coordinates; save console errors. Name files NN_persona_scenario_seq.png so a reviewer can replay your path.

    5. Classify by severity:

    SeverityMeaningExample
    S1 BlockerData loss, dead end, core task impossibleCheckout button dead after retry
    S2 MajorTask possible only via a workaroundError text is a raw stack trace
    S3 MinorWorks but confusing or uglyMisaligned button, unclear label
    S4 NitCosmetic, no task impactTypos, inconsistent spacing
    1. Write the report: for each bug — title, persona + scenario, numbered steps to reproduce, expected vs actual, evidence file names, severity, suggested fix. Rank S1s first, then S2s; merge duplicates; end with "what survived" (flows that withstood hostility) so the team sees wins too.

    Reference

    Hostile moveWhat it exposes
    Submit empty formMissing validation messages
    Back button mid-flowLost state, stale pages
    Refresh after step 2 of 5Non-idempotent steps
    Double-click submitDouble charges, duplicate records
    Paste 1000-char textBroken truncation / overflow
    Append /admin to URLMissing auth redirects

    Evidence housekeeping: one folder per session, files named 01_bob_checkout_doubleclick.png, 02_gina_filter_tap.png, ..., plus a plain-text steps.log listing each action with a timestamp.

    Verification Checklist

    • Ran at least 3

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.

    • Cannot perform deep security penetration or load testing.
    • Requires a live UI: not for pure APIs.
    • Does not replace formal accessibility (a11y) audits.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    6 uses

    Times used to date

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 3 months ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed3 months ago

    What's inside

    Frequently Asked Questions

    It creates up to 3 personas like Boss Bob or Hacker Harry to simulate chaotic, non-linear user behavior.

    No. It focuses on UI/UX resilience and edge-case handling rather than exploitation or infrastructure attacks.

    You get a reproducible log, screenshots of failures, and a bug report ranked by severity (S1 to S4).

    New in Testing & Debugging

    Free