MetaHarness: gaming vertical

A ready-made multi-agent game-design pod scaffolded as a Claude Code harness. One command drops a working claude-code plugin directory on disk with four specialist agents wired up — a concept lead, a mechanics designer, a balance analyst, and a playtest critic — plus the project settings, slash commands, and MCP plumbing needed to actually run them. Aimed at solo designers, game-jam teams, and studios who want a structured pod for ideation and iteration instead of a single generalist prompt. This is a scaffold, not a game engine: it does not ship Unity/Unreal/Godot integrations, no asset pipeline, no live ops — it gives you the design loop, you bring the runtime.

Quickstart

npx @metaharness/gaming@latest my-bot
cd my-bot && npm install && harness doctor

The harness doctor step validates that Claude Code is on PATH, the plugin manifest parses, and each agent's model tier is reachable.

What you get

• agents/concept-lead.md — pitch and high-concept exploration (tier: opus)
• agents/mechanics-designer.md — core loop, systems, and rule specification (tier: sonnet)
• agents/balance-analyst.md — numeric tuning, curves, and economy modelling (tier: sonnet)
• agents/playtest-critic.md — adversarial critique, fun audit, edge-case hunting (tier: haiku)
• commands/pitch.md, commands/loop.md, commands/balance.md, commands/playtest.md — slash commands that route to the right agent
• .claude/settings.json — plugin manifest, permissions, and hook wiring
• CLAUDE.md — pod operating rules: who calls whom, when to fan out, when to converge

Advanced

# Validate the scaffold and the model routing
harness doctor
# => OK  plugin manifest
# => OK  agents/concept-lead.md          (claude-opus)
# => OK  agents/mechanics-designer.md    (claude-sonnet)
# => OK  agents/balance-analyst.md       (claude-sonnet)
# => OK  agents/playtest-critic.md       (claude-haiku)

# Lint frontmatter, agent tier hints, and command routing
harness validate
# => 4 agents, 4 commands, 0 errors

# Run the pod headless against a design brief
claude -p --plugin-dir ./my-bot "/pitch a co-op deckbuilder set on a generation ship"
# => concept-lead drafts pitch -> mechanics-designer proposes core loop
# => balance-analyst flags one runaway resource curve
# => playtest-critic returns 6 fun-risk findings

FAQ

Does this generate a playable game? No. It generates design artifacts — pitch docs, mechanic specs, balance tables, playtest critiques. Hand those to your engine of choice.

Why four agents instead of one prompt? Because "design a game" is the kind of task where a single generalist regresses to genre cliches. Splitting concept / mechanics / balance / playtest forces each pass to defend its choices to the next agent.

Can I swap the model tiers? Yes. Each agent's frontmatter has a model field. Drop playtest-critic to haiku for cheap iteration, bump concept-lead to opus when you want the high-concept work to stretch.

License

MIT. Built on metaharness (https://www.npmjs.com/package/metaharness).

Deep-dive

Full explainer gist: https://gist.github.com/ruvnet/89ab98e6e5f082a376170a941c9d1914