How to use the Isometric Action RPG Builder agent skill

What is an agent skill and what does this one make the agent do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a request matches the skill's description. This skill's description covers game-loop architecture, camera and movement, zones, combat integration, content data, progression, saves, and production-ready vertical slices.

Once active, the agent works in vertical slices. It first aims for one playable loop: title or menu, character choice, movement, one encounter, a reward, progression, and continuation. Further systems are added only after each slice has gameplay proof, automated coverage, and a stable save and load path.

• Inspects the existing game entry point, scene loop, content model, controls, save model, and tests before changing anything.
• Keeps authored content such as classes, equipment, enemies, and zones separate from runtime state.
• Centralises timing, damage, cooldown, and content values instead of duplicating combat clocks across components.
• Consumes authoritative contact events rather than inferring hit outcomes from visual state.
• Adds review query parameters or fixtures for hard states such as boss phase changes, empty inventory, interrupted actions, and save migrations.
• Runs focused unit tests, type or lint checks, a production build, and a real playthrough before release.

Install the build-isometric-arpg skill

Download the skill ZIP from this page. Unzip it so that the build-isometric-arpg folder, containing SKILL.md and agents/openai.yaml, sits inside one of the skills locations below.

For a personal install available across your projects, unzip into ~/.claude/skills/. For a single project, unzip into that project's .claude/skills/ folder so the skill is shared with anyone who works in that repository.

If you use the Claude apps instead of the file system, upload the skill under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/build-isometric-arpg/
• Project: .claude/skills/build-isometric-arpg/ inside your repository

How do you trigger the skill with natural requests?

You do not need a special command. The agent loads the skill when your request matches its description, so name the game type and the stack in plain words. Mention the system you want, such as combat, zones, or saves, so the agent knows which part of the workflow applies.

Describe the outcome you want for the first slice. The skill is built around one playable loop, so a request for a whole game at once will be broken into slices rather than produced in one pass.

• "Start a Three.js isometric action RPG with movement, one melee fight, a loot drop, and a level-up."
• "Add a second zone with a transition gate to my React isometric game and keep saves working."
• "Review my combat code and add a test fixture for the boss phase change."
• "Set up the first vertical slice of an isometric ARPG: title screen, movement, and restart after death."

What do you need before you start?

The skill expects a web project. The source does not name a specific Node.js version, so check what your repository already uses. The guidance assumes a Node.js or similar JavaScript toolchain, because it runs type and lint checks and a production build. The stack should be Three.js, React, or similar web technology.

You also need a real browser for playthrough testing. The skill asks the agent to use the browser the repository approves, and to check keyboard and mouse controls and the intended mobile baseline.

• A web codebase using Three.js, React, or similar technology
• A JavaScript toolchain that supports the repository's type checks, lint, tests, and production build
• A browser for manual or automated playthrough checks
• Version control, so you can review what the agent changed

Safety notes: scripts and network

This skill runs scripts. The agent is expected to edit files, run tests, and run builds in your repository, so review its changes before you merge them. The skill is not meant to change the codebase without review, and it should not be used on projects where that rule applies.

The skill does not use the network, according to its assessment. Scripts it triggers are the repository's own tests, type checks, and build commands rather than downloads from outside.

• Work on a branch and review diffs before merging.
• The skill tells the agent to preserve unmerged work and not build from a stale branch.
• Report baseline failures separately from new failures so you can tell what the agent broke.

Tips for getting good results

Give the agent a clean starting point. The skill asks it to record an authoritative baseline first, so a repository with failing tests makes the agent's report harder to read. Ask it to list baseline failures before it starts.

Ask for one slice at a time and accept it only when it is playable and saves correctly. The skill itself refuses to add a second system before the first has tests and a stable save and load path, so supporting that discipline in your request helps.

• Start with the current codebase rather than a blank folder when you can.
• Ask for test fixtures for tricky states such as boss phases and save migrations.
• Check the fresh start, saved continuation, loss and retry, and one complete objective yourself.
• Keep authored content data separate from runtime state when you review the design.