GDLC Wizard
BaseInfinity/claude-gdlc-wizard
Qualidade de design de jogos em Claude Code via GDLC framework
Ver no GitHubSuite de skills para aplicar Game Development Life Cycle em projetos Claude Code, garantindo que jogos sejam divertidos, justos e bem estruturados — não apenas corretos. Instala ciclos de playtest orientados por personas, detecção de desvios e regras conquistadas automaticamente.
Instalação
npx convez-ai BaseInfinity/claude-gdlc-wizard1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Executar ciclos de playtest estruturados com personas (gameplay, art, pipeline)
- Triangular achados de múltiplos testadores e promover issues críticas
- Manter qualidade de game design ao longo do desenvolvimento iterativo
- Detectar e corrigir desvios de regras de design conquistadas
Veredito

Ferramenta altamente especializada e bem executada para desenvolvimento de jogos em Claude Code. Traz rigor metodológico (GDLC) que diferencia diversão de mero funcionamento técnico — ideal para equipes que levam design de jogos a sério. Instalação idempotente, bem documentada com case studies reais (17+ playtests por jogo).
README
A tradução para o português deste README ainda está na fila.
claude-gdlc-wizard
GDLC enforcement for Claude Code. The sibling of claude-sdlc-wizard, for games instead of code.
SDLC governs whether code is correct. GDLC governs whether a game is fun, fair, readable, and persona-aligned. Most games are correctness-underdetermined — passing every unit test doesn't make them good.
This wizard installs the GDLC skill suite into any Claude Code project: persona-driven playtest cycles, triangulated findings, and a ratchet that only tightens.
Status
Shipped (v0.2.1, 2026-04-25). Phase 1 CERTIFIED; framework playbook consolidated into this repo (Path A); skills migrated to local-repo paths (no sibling clone required). Originally extracted from BaseInfinity/gdlc on 2026-04-23 (now deprecated, see DEPRECATED.md). Case studies to date:
- #1 codeguesser — 17 playtests, 355+ regression tests, 26 earned rules (complete)
- #2 pdlc/TamAGI — consuming now
- #3, #4, #5 — queued (another game project, Canvas-based project, terminal clone)
What You Get
4 skills, installed into your Claude Code project:
| Skill | Invoke | What it does |
|---|---|---|
/gdlc | during a feature cycle | Picks the right playtest cycle type (gameplay-matrix / art-craft-review / pipeline-contract-audit), runs persona agents, triangulates findings, promotes P0s to TDD RED |
/gdlc-setup | first install | Auto-scans your project, detects surfaces (gameplay / art / pipeline), scaffolds a case-study GDLC.md, installs the skill suite. Asks only what scanning can't reveal. |
/gdlc-update | periodically | Reads CHANGELOG, shows rule diffs, runs drift detection on managed files, applies updates selectively while preserving your GDLC.md |
/gdlc-feedback | when you hit a gap | Files structured issues upstream to BaseInfinity/claude-gdlc-wizard — earned-rule candidates, playbook gaps, wizard bugs, methodology questions. Stock GitHub labels. |
Install
Pick whichever fits your environment. All four paths land the same surface: 4 skills + 2 hooks + helper + settings.json + wizard doc. No sibling-repo clone required — the wizard is fully self-contained as of v0.2.1.
Path 1 — npx (recommended)
From your game project root:
npx -y claude-gdlc-wizard init
That installs everything under .claude/ and writes CLAUDE_CODE_GDLC_WIZARD.md to project root. Re-run any time — idempotent. --force overwrites; --dry-run previews.
Path 2 — curl | bash
curl -fsSL https://raw.githubusercontent.com/BaseInfinity/claude-gdlc-wizard/main/install.sh | bash
Wraps the npx flow with Node ≥ 18 preflight + download guard.
Path 3 — Claude Code plugin
Use the Claude Code plugin marketplace mechanism (see Anthropic docs) to install the plugin defined at .claude-plugin/plugin.json. Hooks resolve through ${CLAUDE_PLUGIN_ROOT} automatically.
Path 4 — manual clone
git clone https://github.com/BaseInfinity/claude-gdlc-wizard ~/tmp/claude-gdlc-wizard
node ~/tmp/claude-gdlc-wizard/cli/bin/gdlc-wizard.js init
rm -rf ~/tmp/claude-gdlc-wizard
After install — every path
In Claude Code, run /gdlc-setup. It auto-scans your project, asks the minimum, and scaffolds your GDLC.md. Verify with npx claude-gdlc-wizard check — expect every installed item (settings.json + 3 hook files + 4 skills + wizard doc + .gitignore additions = ~10 rows) to report MATCH.
How It Works
GDLC runs one or more of three cycle types per release, never combined:
| Cycle | Target | Personas | Rubric | Ship Gate |
|---|---|---|---|---|
| Gameplay-matrix | Live game build | 5 gameplay personas (Tourist / Casual / Senior / Purist / Speedrunner — adapt per domain) | Fun / Fair / Readable / Persona-respect / Return | All personas ≥ threshold |
| Art-craft-review | Demo sink (demo.html) | 4 craft personas (Retro Gamer / Designer / Artist / Art Director) | Cohesion / Era / Hierarchy / Craft / Ship | Unanimous Ship |
| Pipeline-contract-audit | Contract harness | Dimension-specific (seed-literate, content-stopword, etc.) | TDD RED before any fix | All RED tests earn GREEN |
Features touching multiple surfaces run cycles sequentially — pipeline first (lock determinism), then gameplay (runtime behavior), then art (visual craft). Never combine rosters across cycles.
Relationship to claude-sdlc-wizard
You should install both. SDLC runs in parallel with GDLC on every feature:
- SDLC — does the code work? Is it tested? (TDD, self-review, CI shepherd)
- GDLC — is the game good? Is it fair? (playtests, triangulation, ratchet)
SDLC without GDLC = clean code that produces boring games. GDLC without SDLC = fun games built with sloppy code.
Contributing
File feedback using /gdlc-feedback from a Claude Code session. It creates a structured GitHub issue on BaseInfinity/claude-gdlc-wizard with stock labels (bug / enhancement / question) and a [<type>] title prefix.
XDLC Ecosystem
| Wizard | Domain | npm | Repo |
|---|---|---|---|
| SDLC | Software engineering | agentic-sdlc-wizard | claude-sdlc-wizard |
| GDLC (this repo) | Game development | claude-gdlc-wizard | claude-gdlc-wizard |
| RDLC | Research | claude-rdlc-wizard | claude-rdlc-wizard |
| Codex SDLC | Codex CLI | codex-sdlc-wizard | codex-sdlc-wizard |
| OpenCode SDLC | OpenCode | opencode-sdlc-wizard | opencode-sdlc-wizard |
License
MIT.
AI Setup Lanes
Three recommended setups in AI_SETUP_LANES.md:
| Lane | Planner | Driver | Reviewer | When |
|---|---|---|---|---|
| A — GDLC Premium | Opus 4.6 max | Opus 4.6 max | GPT-5.5 xhigh | Persona design, playtests, gameplay matrix, architecture |
| B — GDLC Saver | Opus 4.6 max | Sonnet (latest) | GPT-5.5 xhigh | Routine impl, test maintenance, docs |
| C — GDLC Lite | You | Haiku 4.5 | None | Build scripts, asset pipeline, config, grunt work |