How to use the Build Game Map Editor agent skill

What does the build-game-map-editor skill make the agent do?

Once installed, the skill guides the agent through building a paused, top-down director view of your game world. It tells the agent to derive an editor document from your game's real production definitions rather than creating a separate map source of truth.

The playbook covers selection, dragging with snapping, an outliner, layers, camera controls, an inspector for entity settings, enemy aggro, leash and patrol overlays, undo, local draft import and export, and a private editor route protected by an IP allowlist and password sessions.

• Derives a versioned editor document from authored placements, with stable source IDs
• Renders a paused director scene that does not run combat, progression or saves
• Keeps all edits as drafts that only affect the game after a separate reviewed release
• Locks down the editor route with IP allowlists and password sessions

What do you need before installing the skill?

The skill is written for a repository that already contains a Three.js or isometric game with authoritative world and entity definitions. The agent is asked to find those definitions and to use the repository's approved tooling, so the project needs a browser test setup and a release path.

If your game is built in an engine such as Unity or Unreal, this browser-based approach does not apply.

• A repository with a Three.js or isometric game
• Authoritative world and entity definitions in the code
• A browser test setup and a release path that the agent can use

How do you install the skill?

Download the skill ZIP from this page and unzip it. You can install it for yourself or for a single project.

For a personal install, unzip it into your ~/.claude/skills/ folder. For a project install, unzip it into the .claude/skills/ folder inside the project. The skill's main instructions are in SKILL.md, so that file should sit inside the unzipped skill folder.

In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

• Download the ZIP from this page
• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into .claude/skills/ in the project
• Claude apps: upload under Settings → Capabilities → Skills

How do you trigger the skill with a natural request?

You do not need a special command. Describe the editor task in plain language and mention the game and the kind of tool you want. The skill's description lists the kinds of work it covers, so requests that match those tasks are the ones most likely to load it.

Good triggers include asking for a private director view, an outliner or layer panel, enemy aggro or patrol overlays, local draft import and export, or a secured editor route. Say explicitly that you want the editor to avoid changing live gameplay data if that matters to you.

• Ask for a private director view of the level in a Three.js or isometric game
• Ask for an outliner, layers, selection, drag and snap controls
• Ask for enemy aggro, leash or patrol overlays in the inspector
• Ask for local draft import, export, undo and reset
• Ask to secure the editor route with an IP allowlist and password session
• Ask to validate and release the map editor without mutating live gameplay data

What does a typical request look like?

A useful request names the repository, the entities you want to place and the review path you need. For example, you might ask the agent to build a draft-only editor for enemy encounters that shows aggro radius and patrol paths, exports a validated JSON proposal, and never writes to the live game definitions.

Expect the agent to inspect the repository first, locate the authoritative definitions, and then propose the editor structure. The skill tells it to treat export as a proposal, so the result should be a reviewable set of changes rather than an automatic update to the game.

• Example: build a draft-only encounter editor for our Three.js game with enemy aggro and leash overlays
• Example: add undo, redo and local draft storage to the existing map editor, with export validation
• Example: audit our map editor route and check that it fails closed when access settings are missing

What safety limits should you know about?

The skill itself is a set of instructions. According to its listing, it runs no scripts and makes no network requests. The editor it helps you build, however, is a production tool that may be reachable on the web, so the skill's guidance on access control matters.

The skill says a private URL is not enough protection. It recommends failing closed when access configuration is missing, using an exact IP allowlist, signed short-lived session cookies, and keeping passwords and signing material in deployment secrets. It also says the access-gating guidance assumes a private deployment, so public-facing editors need a separate security review.

• The skill listing states it runs no scripts and uses no network
• Passwords, allowlists and signing keys belong in deployment secrets, not in the bundle or logs
• Exports are proposals; they do not change the game until source changes are tested and released
• Do not use the access guidance for a public editor without a security review

What tips help you get the best results?

Start by asking the agent to locate your authoritative world and entity definitions before it designs anything. Keep the editor's data format small and serializable, and keep game-specific adapters outside the reusable editor component.

Expect the skill to insist on validation at every import and export boundary, including schema version, entity count, coordinate limits and matching entity manifests. Ask the agent to test the access-control cases it lists, such as wrong password, wrong IP and logout, before you rely on the editor.

• Point the agent at the real definitions first, so the editor is derived from production data
• Ask for the draft lifecycle to be shown in the UI: production source, editor document, isolated draft, validated export, reviewed integration
• Require confirmation before any reset of draft changes
• Keep flat-plane positions with Y fixed, if your game is flat-world