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.

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)

4 skills, installed into your Claude Code project:

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.

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.

curl -fsSL https://raw.githubusercontent.com/BaseInfinity/claude-gdlc-wizard/main/install.sh | bash

Wraps the npx flow with Node ≥ 18 preflight + download guard.

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.

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

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.

GDLC runs one or more of three cycle types per release, never combined:

Features touching multiple surfaces run cycles sequentially — pipeline first (lock determinism), then gameplay (runtime behavior), then art (visual craft). Never combine rosters across cycles.

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.

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.

MIT.

Three recommended setups in AI_SETUP_LANES.md: