More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Game Engine QA: Spectral Bug Hunt

    1

    A report-first QA protocol for game engines: baseline the exact build, sweep five defect classes, grade every finding with a runtime-verified repro, and gate fixes behind approval.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Game Engine QA: Spectral Bug Hunt

    Game Engine QA: Spectral Bug Hunt

    Example session with this skill installed

    Our 2D game engine is two weeks from a public demo. The unit tests are all green but playtesters keep hitting weird state bugs the tests never catch. Run one of your QA rounds on it and tell us honestly whether it should ship.

    • Read your context and instructions
    • Compiled the game engine qa

    BASELINE: commit 7d3a91c on the release branch, working tree clean. Unit suite 1,482/1,482 green, e2e 31/31 green, production build clean. These are the before-counts every later claim is measured against.

    SPECTRAL SWEEP (built player driven in a real browser runtime, five class passes):

    Crash/hang: one hang repro'd — pausing during a scene transition deadlocks the modal queue; the game accepts no input until reload. State corruption: two findings. Rapid portal re-entry drops carried inventory (repro: enter, exit, re-enter within one second; inventory resets to the entry snapshot). Skin switching mid-load desyncs the loadout panel from the character actually rendered. Contract: the custom-level loader API passes all in-process tests, but in the built player it is unreachable — the bundler tree-shakes the entry point the docs promise, so community levels fail silently. UX integrity: settings changed during a cutscene are discarded without notice. Security hygiene: level files are parsed with a permissive loader that accepts embedded script paths; sandboxed in the demo build, flagged for the full release.

    GRADES AND JUDGMENT: 1 blocker (portal state drop), 2 majors (modal deadlock, loadout desync), 1 major-contract (loader API), 2 minors, 1 nit. Judgment: NOT LAUNCHABLE for the public demo — the blocker corrupts player state in a flow the demo script itself exercises. Report filed, round stopped; no fixes made. Recommended fix-wave scope: the three state-class findings first, each with a fail-before/pass-after regression test, plus a real-runtime e2e spec for the loader contract.

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

    About this skill

    Most QA rounds are wandering with a bug form. This one is a protocol. Phase 0 baselines the exact commit and the before-counts for every suite. Phase 1 sweeps the whole spectrum in defect-class passes — crash and hang, state corruption, contract failures (works in a unit test, unreachable in the real runtime), UX integrity, and security hygiene — driving the built artifact rather than reading source and imagining. Every finding is recorded on a fixed template with a repro that was actually run, a root cause, and an explanation of why existing tests missed it; findings without repros are labeled hypotheses or cut. Phase 2 grades against a severity rubric (blocker / major / minor / nit) and delivers a launch judgment: LAUNCHABLE or NOT LAUNCHABLE, with the reason in one paragraph. Then the round stops. Fixes happen only on explicit approval, on a dedicated branch, aimed at the class of bug rather than the instance, and each ships a regression test proven to fail before the fix and pass after, with prior rounds' repros re-run as a regression sweep before anything is called verified.

    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.

    • Contract-class verification needs a real runtime to drive; without one, those findings are reported as labeled hypotheses, not verdicts.
    • It grades and gates; the decision to launch with known majors stays with your named decision-maker.
    • It is a defect protocol, not a playtesting or game-feel evaluation — fun is a different review.

    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

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    Green suites mostly prove in-process behavior. The protocol hunts the classes suites structurally miss: state corruption under rapid real input, and contract failures where the built artifact differs from the source the tests exercise. Every finding must be repro'd in the real runtime.

    Not during the round. It reports first and stops. Fixes run as a separate fix wave, only on your explicit approval, scoped to the approved findings, each with a regression test proven to fail before and pass after.

    A dated report with the baseline counts, the classes swept, the environments used, and a LAUNCHABLE judgment you can point to — plus the repro library the next round re-runs as its regression sweep.

    New in Testing & Debugging