How to use the Build Rigged Game Assets agent skill

What an agent skill is and what this one does

An agent skill is a folder with a SKILL.md file and optional reference files, scripts and templates that an AI agent reads when a request matches the skill's description. This skill covers the full process for a rigged actor: reading the source, classifying the actor as a character or a monster, building one stable rig, defining actions, wiring the asset into gameplay, and adding catalog media and tests.

The skill tells the agent to inspect the repository before editing anything. It must locate the source model, runtime loader, skeleton conventions, action API, equipment system, collision system, catalog schema and tests, and it must classify delivery as catalog-only, review-only or runtime. An asset is not considered implemented until gameplay actually loads it.

• Characters: a persistent body layer, with swappable equipment such as swords, shields, helmets and armor kept as separate assets bound to named sockets.
• Monsters: one authored combat actor, with separate parts only where they must animate, hide, detach or collide on their own.
• Both: required action roles, navigation and hurt colliders, a manifest validated by a Python script, and deterministic tests.

Install the skill from the ZIP

Download the ZIP for this skill from this page and unzip it. The skill folder must be named build-rigged-game-assets so the agent can find it. You can install it for yourself or for one project.

For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/build-rigged-game-assets/
• Project only: <your-repo>/.claude/skills/build-rigged-game-assets/
• Claude apps: Settings → Capabilities → Skills, then upload the skill.

Trigger the skill with natural requests

You do not need to name the skill. The agent uses it when your request matches the skill description. Describe the actor, the delivery goal and the gameplay you need. Mention the files you have, such as FBX or GLB game actors, and the gear or actions you need.

Good requests name the actor type and the outcome. The skill also covers auditing an existing asset and adding equipment or action controls to a catalog.

• Import this FBX monster and wire it into combat, with telegraph, active and recovery windows for each attack.
• Make a rigged character from this GLB with separate sword, shield and helmet assets, and add a catalog card and inspector.
• Audit the rig and sockets on our existing knight asset and tell me what is missing from the required action roles.
• Add idle, walk, run, attack, hit, dodge, block and death clips to this character and add deterministic tests for them.

Check the manifest with the validator

The skill provides two manifest templates, one for characters and one for monsters. The agent copies the matching template, fills it in and runs the Python validator on it before building. The validator needs Python 3. With the --root option it also checks that shipped files exist in your repository. The --require-verified option is meant only for after every acceptance check has actually passed.

You can run the validator yourself to see what the agent is checking.

• Templates: assets/character-asset-manifest.template.json and assets/monster-asset-manifest.template.json
• Validator: scripts/validate_asset_manifest.py

Review the browser checks the agent will run

The skill asks for browser verification in the repository-approved browser. The catalog card must be a transparent PNG with zero WebGL canvases. The selected inspector must hold exactly one live model canvas. The agent is asked to click every action, toggle every character slot and all equipment off, inspect the model from front, back and both sides, and check the narrow or mobile layout.

Read these checks before you accept a result. The skill states that a screenshot or a separate look-dev mesh is not runtime proof.

• Card: one transparent image and zero canvases.
• Inspector: one model host and canvas, every action playable, slots toggled cleanly.
• Failure paths: fallback image and model-load failure behavior are tested.
• Console: no broken media and no warnings or errors.

Keep the commit narrow and reportable

The skill asks the agent to commit only the narrow asset lane and to report the manifest, main and runtime models, gear files, rig and sockets, actions, catalog route, status, verification results, known boundaries and the exact commit SHA. Full-suite baseline failures must be reported separately from the asset's own results.

Before committing, the agent runs the project's focused asset tests, the manifest validator, the build, lint, the full test suite and git diff --check. The exact commands depend on your project.