agentleFS
xsollaxsolla/xsolla-ai-kit/.cursor/skills/nocode-shop-setup/SKILL.md

nocode-shop-setup

Skill9 starsChanged 5 months agoMITView on GitHub

Builds a shop, web shop or storefront on Xsolla's side with Shop Builder: no code, servers or deploys. Use when the user says build me a shop or "create a web shop for my game", unless the user will code it, and for "I don't want to write code". No local repo or no coding means this path. Also installs the Xsolla CLI; the user signs in. If the user will write and deploy code, open shop-setup.

  • Reads credentials

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 5 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run these first, so a blocker shows up in the first message, not after the catalog is built.

    line 37
  2. Distinctive:

    The game: Read the user's store listing link, if they gave one: name, platforms, developer, description, icon and screenshots.

    line 71
  3. Very specific:

    The catalog, LiveOps and Login projects: Read the project's catalog groups, items, bundles and LiveOps with the read-only xsolla catalog …

    line 73
  4. Very specific:

    Do not write the answer to .env.

    line 97
  5. Distinctive:

    When a Shop Builder site, new or existing, has to look like a Figma design, follow Matching a Figma design in Shop Builder: standard blocks …

    line 101
  6. Distinctive:

    Ask everything still missing in one message, each question with a recommended answer, so a reply such as "all recommended" settles it:

    line 111
  7. Very specific:

    Ask in plain words, with one filled-in example per answer.

    line 131
  8. Distinctive:

    Run these before the storefront.

    line 136
  9. Very specific:

    Do not start description-to-shop or shop-builder-assembly for a game home.

    line 161
  10. Very specific:

    Do not start game-web-portal for a store.

    line 162
  11. Very specific:

    Before showing the proposal, check it against the limits in catalog-design's LiveOps reference: each offer chain or reward chain step takes …

    line 175
  12. Very specific:

    When the batch said a game server receives paid orders, run webhooks-impl.

    line 189
  13. Very specific:

    Do not run production here.

    line 193

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

No-code shop setup

Build a site in Shop Builder. The agent sends commands; the site is created there. The user arranges blocks, configures them, and publishes. Payments use Pay Station. Custom blocks are possible, and that code still runs on Xsolla's side.

The user acts three times: the build request, one batch of answers (only when a value is still missing), and one yes. When the Xsolla CLI isn't signed in yet, signing in, in their own terminal, comes before the batch. Everything else runs without them. Never stop to ask "should I continue?".

Which side

  • The user will write and deploy the code → stop and open shop-setup.
  • Otherwise, no local repo, or the user says they don't code or don't want to run servers → Xsolla's side. Say so in one line and continue; don't ask.
  • A local repo with storefront code and no stated preference → put the question in the batch below, with Shop Builder as the recommended answer when the user has not said they will maintain the code. Shop building paths in AGENTS.md explains the two options.

Check what can be built, before any question

Run these first, so a blocker shows up in the first message, not after the catalog is built.

  1. The CLI: installed, signed in, current. This path runs only through the Xsolla CLI. If xsolla --version fails, say so in one line and install it now; on Windows, too. Then the user runs xsolla auth login in their own terminal: it opens a browser and doesn't complete from the agent's shell. Steps per platform: CLI setup. Never ask for a pasted API key, or for a Publisher Account link to make one, instead. Then run xsolla update --check; when it can't tell (it has reported "not published") and the CLI came from Homebrew, run brew upgrade xsolla. Refresh the kit with xsolla skills install --global --source ai-kit. Name both versions in the first message. The checks below are read-only, except the CLI switch in check 2.
  2. The project and sandbox mode. Take the merchant and project from the Publisher Account link in the request, else from xsolla config list --json, and don't ask to confirm it: the plan names the project. With neither, ask for the project link in the batch. Shop Builder needs CLI sandbox mode off, and the backup and preflight refuse sandbox mode, so this can't wait for the plan. When the CLI points at another project or has sandbox on, switch it yourself (xsolla config set merchant-id <id>, xsolla config set project-id <id>, xsolla config set sandbox false). Say in the first message that the CLI config keeps it, so every later xsolla command, in this session and others, reaches the live project, payments and refunds included, and give the undo: xsolla config set sandbox true and the previous merchant and project, by ID. Nothing is written to the project before the plan's yes.
  3. The CLI can make a site that opens. A new store site gets its common layout only from a CLI with xsolla shopbuilder create-block. Run xsolla shopbuilder create-block --help. If the usage line doesn't name create-block, say in the first message: a new site this CLI creates won't open in the Editor, Publish or private preview until a newer CLI ships. Put the choice in the batch. Recommended: the developer creates the site in Publisher Account (Storefronts → Websites, from the Single-page shop template) and gives its address; assembly builds on it, with more pages when the preset has them, and it opens in the Editor and preview. The other answers: build the structure anyway, or wait. A game home is built on the partner's template site; see Store or game home.
  4. The game. Read the user's store listing link, if they gave one: name, platforms, developer, description, icon and screenshots.
  5. The catalog, LiveOps and Login projects. Read the project's catalog groups, items, bundles and LiveOps with the read-only xsolla catalog and xsolla liveops list commands, and its Login projects. Draft catalog-design's proposal now, so its questions go in the batch instead of a second round.
  6. Browser and design tools in this session. Find the browser that is signed in to Publisher Account (a built-in browser, Claude in Chrome, or another extension) and use it for every browser step; don't ask which one. Check that the mockup's design tool, such as Figma, is available in this session: a setup log from an earlier session is not proof. When one is missing, say in the first message what that changes: no visual review without a signed-in browser, an HTML mockup instead of Figma.

Store or game home

  • A game home is a Game Web Portal: a home around the store, with news, rewards and community, for a game on PC, the App Store or Google Play. Only a PC game gets the optional launcher section.
  • An existing site with pages /news, /rewards, and /store is a game home. The partner created it from the Multi-page web portal template. A site without those pages is a store.
  • Otherwise, unless the request already says, put store or game home in the batch, with store as the recommended answer.
  • For a game home, the first message says to create the site in Publisher Account if it doesn't exist yet: Storefronts → Websites → Multi-page web portal → Manual.

Do not write the answer to .env.

A Figma design

When a Shop Builder site, new or existing, has to look like a Figma design, follow Matching a Figma design in Shop Builder: standard blocks first; a Custom Block (xsolla shopbuilder create-custom-block) for each section a standard block can't match one to one, named in the plan by its purpose with the reason; never custom code, and never touch custom code already on the site; prices, timers, items, cart and login stay live; never publish; check at 1440, 1024 and 390 px against the Figma frames, and list what neither can match.

One batch of questions

Ask everything still missing in one message, each question with a recommended answer, so a reply such as "all recommended" settles it:

  • which side, only when the rule above says to ask;
  • the project link, only when neither the request nor the CLI names a project;
  • store or game home;
  • build without the common layout or wait, only when the CLI lacks create-block;
  • the storefront intake the listing didn't answer: game name, platforms, a short description, the site's address, and its languages (recommended: the primary language only; more go in later with shopbuilder-translate, which shows its own preview). The storefront skills read these answers and don't ask them again;
  • catalog-design's questions: the archetype, what to do with test or demo entries already in the project (recommended: leave them and show only the new groups on the site; disabling one changes the live catalog at once), and any commercial value it can't infer;
  • item delivery: whether a game server should receive paid orders. Recommended: not now; purchases still work, but items reach the game only once the developer deploys webhook handler code on their own server;
  • what merchant-setup still needs after reading the project.

Ask in plain words, with one filled-in example per answer. Never send a template with placeholders such as <name> or <link> for the user to fill in.

Before the site

Run these before the storefront. Do not run Headless Checkout or the headless login widget: this site renders its own checkout and login.

  1. merchant-setup — Publisher Account, project, and API key. The CLI signed in at check 1 holds the project key: don't ask the user for one, in the batch or later, and don't take that skill's pasted-key route: this path has none. Creating a project key is a write: never before the yes. If the CLI's key fails a read, put creating one in the batch.
  2. catalog-design — the items, packs, bundles and LiveOps the site will sell, through the CLI's catalog and liveops commands. Only its catalog steps: the archetype, the proposal and pricing. No client catalog, purchase code or checkout: the site has its own. If the project already has that catalog, do not rebuild it. Its proposal waits for the one yes below.
  3. login-setup — only for the Login project the site signs players in with. Before the yes, only read the existing ones. Creating or linking one is a write: put it in the one approval's proposal, or list it as a Publisher Account step. No Login Widget code and no login-styling.

What to run

  • Store, and the input is a description or a store listing link → description-to-shop. It writes the brief and then calls shop-builder-assembly.
  • Store, and a validated brief already exists → shop-builder-assembly.
  • Game home → game-web-portal.

Do not start description-to-shop or shop-builder-assembly for a game home. Do not start game-web-portal for a store.

One approval

Show the catalog proposal and the storefront skill's plan in one message, with a desktop and mobile mockup for a store (a game home shows game-web-portal's plan), and ask for one yes. That yes lets catalog-design create the catalog and the storefront skill apply its plan, in that order. The storefront skill still runs its own preflight and backup, and asks again only for what its plan didn't show. When the catalog already exists, the plan alone is the message. When the proposal writes to the catalog, the plan says so: the project is live, and catalog changes reach the store and game clients at once, with no publish step.

Before showing the proposal, check it against the limits in catalog-design's LiveOps reference: each offer chain or reward chain step takes one item, and offer chain steps only items and bundles. Include every value the catalog API requires, such as the virtual currency's own unit price, so none has to be invented later. After the yes, if something can't be created exactly as proposed, create the rest, then show what differs and ask before substituting anything: an extra bundle or a changed step is a change the yes didn't cover.

The batch above asks the storefront intake. With a project directory, record the answers in the onboarding ledger; without one, they stay in the conversation. The storefront skills read them and ask only about a conflict. Do not store the store-or-game-home choice in .env.

After the site

When the batch said a game server receives paid orders, run webhooks-impl. It writes handler code the developer deploys on their own server; say so. Otherwise skip it and list item delivery as not connected in the handoff.

Do not run production here. Leaving sandbox on a shop the user deploys is shop-setup.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.