Ad-Driven Game Skills

    1

    33 MIT-licensed AI skills for designing, building and improving Android and iOS games—from gameplay and progression to rewarded ads, analytics and release.

    Free

    3 usesSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    About this skill

    A practical, MIT-licensed skill pack for AI agents working on Android and iOS games. The v0.2.0 library includes 33 focused skills covering game design, prototyping, UI and onboarding, progression, rewarded ads, purchases, saves, analytics, release and live operations.

    Start with ad-game-director for broad planning or an existing-game audit. Choose a specialist when the task is focused. Each skill includes its own resources and licence and can also be installed separately from the public repository.

    Included: actionable workflows, project and output templates, synthetic examples, behavioural evaluation records, and local Python helpers for economy, retention cohorts, revenue sensitivity, experiment readouts, offline accrual, contrast and artifact size budgets.

    The core guidance is engine-independent, with detailed Godot, AdMob and Firebase integration recipes. Recipes describe contracts and verification steps; they are not a bundled production SDK plugin or complete game starter kit.

    The library separates inspected source, tests, simulations, proposed designs and live measurement. It makes no claim of measured retention or revenue uplift. Validate integrations on the actual game, SDK versions and devices before release.

    Source and installation: GitHub repository and README. Free to use and adapt under the MIT licence. Linked third-party documentation retains its own terms. Original library validation and 125 unit tests passed again on 27 October 2026; those checks do not establish live game or provider behaviour. Agensi-installed agent discovery has not been smoke tested.

    Example: a bounded prototype plan

    Request

    Use Ad-Driven Game Skills to plan a two-week Godot prototype of a portrait idle tower-defence game for Android and iOS. One developer, no live player or ad data. Keep a playable free loop and optional rewarded ads. Give me a bounded plan, reward-reliability acceptance checks and the first playtest questions. Treat all numbers as design assumptions.

    Response generated from the library guidance

    Generated planning example — proposed work, not an implemented or tested game. All quantities below are design assumptions for this prototype. No live player, advertising or commercial results are available.

    Build a portrait tower-defence slice that tests one question: Do placement and upgrade choices remain understandable and interesting while combat runs automatically?

    Use two weeks as ten working days for one developer. Keep the project's selected Godot version; verify its export requirements before implementation. Intended audience, target devices and art direction remain open decisions.

    Playable scope

    Assume one map, eight waves, two tower types and three enemy types. Use placeholder art with clear silhouettes and threat feedback. The player places defences, observes automatic combat and chooses between damage and slowing upgrades.

    The complete loop is start → place → defend → win or lose → understand the result → buy a persistent upgrade → restart. A loss should explain an actionable mistake. Keep the free route complete: no waiting for an ad, mandatory revival or purchased resource is needed to play again.

    For this slice, “idle” means automatic combat during active play. Backgrounding pauses combat; offline earnings, prestige, cloud saves, purchases, leaderboards and notifications are outside scope.

    Assume a single optional post-run reward: 25% extra research earned in that run, rounded down. Calculate and snapshot the exact amount before offering it. Show “Claim bonus” and a clear skip option; skipping preserves ordinary earned research. Use a fake ad adapter with controlled outcomes during prototype development.

    Ten-day work plan

    DaysDeliverable and check
    1–2Define rules, audience assumptions, scene boundaries and tunable values. Implement one reproducible wave with meaningful placement choices.
    3–4Complete the eight-wave run, outcome explanation and immediate restart. Check fresh-player and deliberately weak strategies.
    5–6Add versioned local saves, two persistent upgrades and the fake rewarded-offer flow. Compare no-ad and bonus-assisted progression using explicit fixtures.
    7–8Exercise reward and save failures. Check touch controls, pause/resume and repeated runs on available Android/iOS devices. Record untested exports or device paths honestly.
    9–10Run authorised playtests, revise the largest comprehension problem and repeat the relevant task. Deliver build instructions, observations and a continue/revise/stop recommendation.

    If core combat is still unclear after day four, reduce content before adding progression.

    Reward-reliability acceptance checks

    • Ad unavailable, declined or failed: ordinary rewards and restart still work.
    • Load or retry succeeds: an ad still requires the player's explicit choice.
    • Close without earn: no bonus.
    • Duplicate earn callbacks: one durable grant.
    • Valid earn after close or navigation: credit the original run's immutable claim once.
    • Persisted earned receipt followed by grant-save failure: retain pending settlement and retry without replaying an ad.
    • Grant committed, then crash before acknowledgement: restart finds the claim already applied and does not duplicate it.
    • Save the reward and deduplication marker together; do not treat two unrelated writes as atomic.
    • Stale load callbacks cannot restore eligibility after a privacy-state change; legitimate earned rewards remain settleable.

    Receipt persistence can itself fail. If no recoverable receipt exists before a crash, local recovery is not guaranteed; expose the error instead of claiming success. Fake-ad checks establish state handling only. Real SDK playback, privacy flows and native callback behaviour require later verification.

    First playtest tasks and questions

    Begin without coaching: “Protect the path and prepare for a tougher wave.”

    Observe whether players can identify the next action, explain why a wave failed, change strategy and find the next attainable upgrade. Record assistance separately.

    Ask:

    • “What did you expect that tower or upgrade to do?”
    • “What caused that result?”
    • “What would you change on another run?”
    • “What does the optional bonus give you?”
    • “What happens if you skip it?”
    • “Would you like another run, or would you stop here—and why?”

    Use these observations to decide whether the slice teaches its choices and warrants another prototype. Preference and voluntary replay in a small study do not establish retention or revenue.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Game Development.

    Guidance, templates and local tools; no complete game starter or production native SDK plugin. No measured retention or revenue uplift. Native device/provider integrations need project-specific verification and current official documentation. Codex discovery was smoke tested; Claude Code/Cursor were documentation checked. Agensi installation/discovery is untested.

    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

    3 uses

    Times used to date

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 21 days ago

    • Free with an Agensi account

    Nothing to install. It runs in your AI agent.

    Optional
    Python 3.9+Git

    Python runs the calculators and installer. Git installs from source and gets updates.

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Browser

    Read access is used for selected bundled resources and project context. Shell and write access are task-dependent for local helpers, implementation and output artifacts. Browser/network access is used to verify current official documentation or user-authorised integrations. Local calculators do not contact ad networks or send telemetry. No external account is universally required. Scope project files and provider access to the user's actual request.

    Listed21 days ago

    What's inside

    Frequently Asked Questions

    The v0.2.0 pack contains 33 focused skills, including Ad Game Director, with standalone resources and MIT licence copies. It includes project templates, synthetic fixtures, local Python helpers, documentation and validation records.

    The repository documents installation for Codex, Claude Code and Cursor. Codex CLI discovery was smoke tested; Claude Code and Cursor compatibility was documentation checked. Agensi-installed discovery and other clients have not been smoke tested.

    The core design and decision guidance is engine-independent. Godot, AdMob and Firebase have detailed integration recipes. The pack does not include a production native SDK plugin or complete game starter kit.

    No. Design ideas and economic assumptions need project-specific testing and real cohorts. Inspected source, tests, simulations and live measurements are kept distinct; the library claims no measured retention or revenue uplift.

    Start with ad-game-director for a broad planning or audit request, or name a specialist for a focused task. Use the linked repository README for installation and selective profiles. Each skill folder can also be copied alone with its resources and licence; the optional installer preserves local edits.

    The agent needs to read the selected skill files. Python 3.9+ is required for local helpers and the optional installer; Git is optional for source installation and updates. Game implementation, current documentation checks and provider work use task-specific tools and access. No external provider account is universally required, and local calculators do not send telemetry.

    New in Game Development

    Free