gstack-game
Game development workflow skills for Claude Code. 29 interactive skills for game development — strongest in design review and planning, with dev-phase support. When you notice the user is at these stages, suggest the appropriate skill: - User has a fragile mood/image/mechanic fragment and wants creative sparks, not critique → suggest /spark-lens - User has a PDF/doc/notes they want to turn into a GDD → suggest /game-import - Brainstorming a game idea → suggest /game-ideation - Reviewing a game plan (strategy/direction) → suggest /game-direction - Reviewing a game design document → suggest /game-review - Reviewing technical architecture for a game → suggest /game-eng-review - Reviewing game economy or balance → suggest /balance-review - Simulating player experience → suggest /player-experience - Reviewing game UI/UX → suggest /game-ux-review - Evaluating a game pitch or proposal → suggest /pitch-review - Code review on a game project → suggest /gameplay-implementation-review - QA testing a game → suggest /game-qa - Ready to ship / create PR for a game → suggest /game-ship - Debugging a game bug → suggest /game-debug - Sprint/milestone retrospective → suggest /game-retro - Wanting adversarial second opinion → suggest /game-codex - Post-release doc updates → suggest /game-docs - Visual quality audit → suggest /game-visual-qa - Reviewing game assets (art/audio pipeline) → suggest /asset-review - Designing a playtest protocol → suggest /playtest - Working with production or live systems → suggest /careful - Want to scope edits to one directory → suggest /guard - Removing edit restrictions → suggest /unfreeze If the user pushes back on skill suggestions ("stop suggesting", "too aggressive"): 1. Stop suggesting for the rest of this session 2. Run: gstack-config set proactive false 3. Say: "Got it — I'll stop suggesting skills. Say 'be proactive' to turn them back on." If the user says "be proactive again" or "turn on suggestions": 1. Run: gstack-config set proactive true 2. Say: "Proactive suggestions are back on."
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/fagemx--gstack-game)HTML
<a href="/listings/fagemx--gstack-game"><img src="/api/listings/fagemx--gstack-game/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
1. Install gstack-game (30 seconds — see below) 2. Run `/triage` — it detects your project state and tells you which skill to run first. 3. Or jump straight in: `/spark-lens` (have a fragile fragment), `/game-import` (have a doc), `/game-ideation` (ready to structure), `/game-review` (have a GDD). 4. Most structured skills end with a **Next Step** recommendation — follow the chain when you are ready. 5. Stop after 2-3 skills. You'll know if this is for you.
---
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
SKILL.md · Manifest · 229 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.
- ---
- name: gstack-game
- version: 0.5.0
- description: |
- Game development workflow skills for Claude Code. 29 interactive skills
- for game development — strongest in design review and planning, with dev-phase support.
- When you notice the user is at these stages, suggest the appropriate skill:
- - User has a fragile mood/image/mechanic fragment and wants creative sparks, not critique → suggest /spark-lens
- - User has a PDF/doc/notes they want to turn into a GDD → suggest /game-import
- - Brainstorming a game idea → suggest /game-ideation
- - Reviewing a game plan (strategy/direction) → suggest /game-direction
- - Reviewing a game design document → suggest /game-review
- - Reviewing technical architecture for a game → suggest /game-eng-review
- - Reviewing game economy or balance → suggest /balance-review
- - Simulating player experience → suggest /player-experience
- - Reviewing game UI/UX → suggest /game-ux-review
- - Evaluating a game pitch or proposal → suggest /pitch-review
- - Code review on a game project → suggest /gameplay-implementation-review
- - QA testing a game → suggest /game-qa
- - Ready to ship / create PR for a game → suggest /game-ship
- - Debugging a game bug → suggest /game-debug
- - Sprint/milestone retrospective → suggest /game-retro
- - Wanting adversarial second opinion → suggest /game-codex
- - Post-release doc updates → suggest /game-docs
- - Visual quality audit → suggest /game-visual-qa
- - Reviewing game assets (art/audio pipeline) → suggest /asset-review
- - Designing a playtest protocol → suggest /playtest
- - Working with production or live systems → suggest /careful
- - Want to scope edits to one directory → suggest /guard
- - Removing edit restrictions → suggest /unfreeze
- If the user pushes back on skill suggestions ("stop suggesting", "too aggressive"):
- 1. Stop suggesting for the rest of this session
- 2. Run: gstack-config set proactive false
- 3. Say: "Got it — I'll stop suggesting skills. Say 'be proactive' to turn them back on."
- If the user says "be proactive again" or "turn on suggestions":
- 1. Run: gstack-config set proactive true
- 2. Say: "Proactive suggestions are back on."
- allowed-tools:
- - Bash
- - Read
- - AskUserQuestion
- ---
- <!-- AUTO-GENERATED from SKILL.md.tmpl — do not edit directly -->
- <!-- Regenerate: bun run build -->
- ## Preamble (run first)
- ```bash
- _GD_VERSION="0.5.0"
- _GG_BIN=""
- for _p in "bin" ".claude/skills/gstack-game/bin" "$(dirname "$(readlink -f .claude/skills/game-review/SKILL.md 2>/dev/null)" 2>/dev/null)/../../bin"; do
- [ -f "$_p/gstack-config" ] && _GG_BIN="$_p" && break
- done
- [ -z "$_GG_BIN" ] && [ -f ".claude/skills/gstack-game/bin/gstack-config" ] && _GG_BIN=".claude/skills/gstack-game/bin"
- [ -z "$_GG_BIN" ] && echo "WARN: gstack-game bin/ not found, some features disabled"
- mkdir -p ~/.gstack/sessions
- touch ~/.gstack/sessions/"$PPID"
- _SESSIONS=$(find ~/.gstack/sessions -mmin -120 -type f 2>/dev/null | wc -l | tr -d ' ')
- find ~/.gstack/sessions -mmin +120 -type f -delete 2>/dev/null || true
- _PROACTIVE=$([ -n "$_GG_BIN" ] && "$_GG_BIN/gstack-config" get proactive 2>/dev/null || echo "true")
- _BRANCH=$(git branch --show-current 2>/dev/null || echo "unknown")
- _TEL_START=$(date +%s)
- _SESSION_ID="$$-$(date +%s)"
- _LAKE_SEEN=$([ -f ~/.gstack/.completeness-intro-seen ] && echo "yes" || echo "no")
- _TEL_PROMPTED=$([ -f ~/.gstack/.telemetry-prompted ] && echo "yes" || echo "no")
- mkdir -p ~/.gstack/analytics
- echo '{"skill":"gstack-game","ts":"'$(date -u +%Y-%m-%dT%H:%M:%SZ)'","repo":"'$(basename "$(git rev-parse --show-toplevel 2>/dev/null)" 2>/dev/null || echo "unknown")'"}' >> ~/.gstack/analytics/skill-usage.jsonl 2>/dev/null || true
- echo "BRANCH: $_BRANCH"
- echo "PROACTIVE: $_PROACTIVE"
- echo "GD_VERSION: $_GD_VERSION"
- echo "SESSIONS: $_SESSIONS"
- echo "LAKE_INTRO: $_LAKE_SEEN"
- echo "TEL_PROMPTED: $_TEL_PROMPTED"
- ```
- If `PROACTIVE` is `"false"`, do NOT proactively suggest gstack-game skills — only invoke
- them when the user explicitly asks. The user opted out of proactive suggestions.
- If `LAKE_INTRO` is `no`: Before continuing, introduce the Completeness Principle.
- Tell the user: "gstack-game follows the **Boil the Lake** principle — always do the complete
- thing when AI makes the marginal cost near-zero. In game dev: don't skip balance data
- ('we'll tune later'), don't skip FTUE design ('tutorial can wait'), don't skip controller
- support ('polish later'). These are lakes, not oceans."
- Then:
- ```bash
- touch ~/.gstack/.completeness-intro-seen
- ```
- This only happens once.
- If `TEL_PROMPTED` is `no` AND `LAKE_INTRO` is `yes`: Ask the user about telemetry.
- Use AskUserQuestion:
- > Help gstack-game improve! Anonymous mode tracks which skills you use and how long
- > they take — no code, file paths, or repo names are ever recorded.
- > Change anytime with `gstack-config set telemetry off`.
- Options:
- - A) Sure, anonymous is fine (recommended)
- - B) No thanks, fully off
- If A: run `gstack-config set telemetry anonymous`
- If B: run `gstack-config set telemetry off`
- Always run:
- ```bash
- touch ~/.gstack/.telemetry-prompted
- ```
- This only happens once.
- ## Session Awareness
- If `SESSIONS` is `3` or more: The user has 3+ concurrent Claude Code windows.
- For EVERY AskUserQuestion in this session, prepend a re-ground line:
- > **[{repo} / {branch}]** You're working on {current task}.
- This prevents context confusion when switching between windows.
- ## AskUserQuestion Format (Game Design)
- **ALWAYS follow this structure for every AskUserQuestion call:**
- 1. **Re-ground:** Project, branch, what game/feature is being reviewed. (1-2 sentences)
- 2. **Simplify:** Plain language a smart 16-year-old gamer could follow. Use game examples they'd know (Minecraft, Genshin, Among Us, Celeste, Hades) as analogies.
- 3. **Recommend:** `RECOMMENDATION: Choose [X] because [one-line reason]` — include `Player Impact: X/10` for each option. Calibration: 10 = fundamentally changes player experience, 7 = noticeable improvement, 3 = cosmetic/marginal.
- 4. **Options:** Lettered: `A) ... B) ... C) ...` with effort estimates (human: ~X / CC: ~Y).
- **Game-specific vocabulary — USE these terms, don't reinvent:**
- - Core loop, session loop, meta loop
- - FTUE (First Time User Experience), aha moment, churn point
- - Retention hook (D1, D7, D30)
- - Economy: sink, faucet, currency, exchange rate
- - Progression: skill gate, content gate, time gate
- - Bartle types: Achiever, Explorer, Socializer, Killer
- - Difficulty curve, flow state, friction point
- - Whale, dolphin, minnow (spending tiers)
- ## Completeness Principle — Boil the Lake
- AI-assisted development makes the marginal cost of completeness near-zero.
- - If Option A is complete and Option B is a shortcut — **always recommend A**.
- - **Lake:** Full balance spreadsheet, complete FTUE flow, all input methods, regression tests.
- - **Ocean:** Rewriting the engine, porting platforms from scratch, building custom MMO networking.
- **When estimating effort**, show both scales:
- | Task type | Human team | CC+gstack-game | Compression |
- |-----------|-----------|----------------|-------------|
- | GDD section drafting | 2 days | 30 min | ~50x |
- | Balance spreadsheet | 1 day | 15 min | ~50x |
- | Playtest protocol | 4 hours | 15 min | ~15x |
- | Economy model review | 1 day | 30 min | ~30x |
- | Code review (gameplay) | 2 hours | 10 min | ~12x |
- | Architecture design | 2 days | 4 hours | ~5x |
- **Anti-patterns:**
- - "We'll balance it later." (Balance data is the cheapest lake.)
- - "Tutorial can wait." (FTUE determines D1 retention.)
- - "Skip controller support." (If target platform, input is a lake.)
- - "Tests are overkill for a game." (Gameplay regression tests catch silent balance breaks.)
- ## Search Before Building
- Before building any game system, search first.
- **Three layers:**
- - **Layer 1** (tried and true): ECS, A*, behavior trees, sink/faucet. Don't reinvent.
- - **Layer 2** (new and popular): Latest networking framework, procedural gen. Scrutinize.
- - **Layer 3** (first principles): "What if death IS the progression?" — prize these above all.
- **Eureka moment:** When first-principles reasoning challenges conventional game design:
- ```bash
- jq -n --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" --arg skill "SKILL_NAME" --arg branch "$(git branch --show-current 2>/dev/null)" --arg insight "ONE_LINE_SUMMARY" '{ts:$ts,skill:$skill,branch:$branch,insight:$insight}' >> ~/.gstack/analytics/eureka.jsonl 2>/dev/null || true
- ```
- ## Completion Status Protocol
- - **DONE** — All steps completed. Evidence provided.
- - **DONE_WITH_CONCERNS** — Completed, with issues listed.
- - **BLOCKED** — Cannot proceed. State what's blocking.
- - **NEEDS_CONTEXT** — Missing information. State what you need.
- Escalation: 3 failed attempts → STOP. Security/balance uncertainty → STOP. Bad work is worse than no work.
- ## Skill Routing
- | User is doing... | Suggest |
- |-------------------|---------|
- | "I have a weird fragment / mood / image" | `/spark-lens` |
- | "I have a GDD/PDF/doc to import" | `/game-import` |
- | "I have a game idea" | `/game-ideation` |
- | "Should we build this?" | `/game-direction` |
- | "Review this GDD" | `/game-review` |
- | "Is the architecture right?" | `/game-eng-review` |
- | "Is the economy balanced?" | `/balance-review` |
- | "What does the player experience?" | `/player-experience` |
- | "Is the UI good?" | `/game-ux-review` |
- | "Will this pitch work?" | `/pitch-review` |
- | "Review this PR/code" | `/gameplay-implementation-review` |
- | "Test this build" | `/game-qa` |
- | "Ship this" | `/game-ship` |
- | "Why is this broken?" | `/game-debug` |
- | "How was this sprint?" | `/game-retro` |
- | "Find holes in this" | `/game-codex` |
- | "Update the docs" | `/game-docs` |
- | "Does this look right?" | `/game-visual-qa` |
- | "Check these assets" | `/asset-review` |
- | "Plan a playtest" | `/playtest` |
- | "Be careful" | `/careful` |
- | "Lock edits to this folder" | `/guard` |
- | "Unlock editing" | `/unfreeze` |
- ## Telemetry (run last)
- ```bash
- _TEL_END=$(date +%s)
- _TEL_DUR=$(( _TEL_END - _TEL_START ))
- [ -n "$_GG_BIN" ] && "$_GG_BIN/gstack-telemetry-log" \
- --skill "gstack-game" --duration "$_TEL_DUR" --outcome "OUTCOME" \
- --used-browse "false" --session-id "$_SESSION_ID" 2>/dev/null &
- ```
README.md · README · 356 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.
- # gstack-game
- **Complete game development workflow skills for Claude Code**
- [繁體中文](README.zh-TW.md)
- 29 interactive AI skills for game development — from creative spark to shipped build. Covers the **complete game production workflow**: spark exploration, design review, prototype planning, implementation handoff, game feel diagnosis, playability assessment, code review, QA, and release. Built on [gstack](https://github.com/garrytan/gstack)'s engineering architecture and review methodology, fully rewritten for game development.
- > **What this IS:** A structured game development support system — it helps you protect early creative sparks, then judge, score, and improve your game design and code when you are ready.
- > **What this is NOT:** A game builder or code generator — it doesn't implement features or create assets for you.
- gstack is Garry Tan's AI engineering workflow for Web/SaaS. gstack-game ports that same methodology to game dev: game design theory (MDA, SDT, Flow State) replaces SaaS metrics (MRR, churn). Core loop, retention hooks, and Sink/Faucet economy models replace API endpoints and database schemas. The engineering backbone (template engine, preamble injection, anti-sycophancy) maintains gstack-level quality.
- **Who this is for:**
- - **Solo indie developers** — structured design review and QA even as a one-person team
- - **Small teams (2-5)** — AI skills fill missing specialist roles (economy designer, UX researcher, QA lead)
- - **Game design students** — structured design thinking framework with game design theory built into every skill
- ---
- ## Quick start: your first 10 minutes
- 1. Install gstack-game (30 seconds — see below)
- 2. Run `/triage` — it detects your project state and tells you which skill to run first.
- 3. Or jump straight in: `/spark-lens` (have a fragile fragment), `/game-import` (have a doc), `/game-ideation` (ready to structure), `/game-review` (have a GDD).
- 4. Most structured skills end with a **Next Step** recommendation — follow the chain when you are ready.
- 5. Stop after 2-3 skills. You'll know if this is for you.
- ---
- ## Install — 30 seconds
- **Requirements:** [Claude Code](https://docs.anthropic.com/en/docs/claude-code), [Git](https://git-scm.com/), [Bun](https://bun.sh/) v1.0+
- ### Option A: Paste to Claude (recommended)
- Open Claude Code and paste this. Claude does the rest.
- > Install gstack-game: run **`git clone https://github.com/fagemx/gstack-game.git ~/.claude/skills/gstack-game && cd ~/.claude/skills/gstack-game && bun run build`** then add a "gstack-game" section to CLAUDE.md that lists the available skills: /triage, /spark-lens, /game-import, /game-ideation, /game-direction, /game-review, /game-eng-review, /balance-review, /player-experience, /game-ux-review, /plan-design-review, /pitch-review, /prototype-slice-plan, /implementation-handoff, /gameplay-implementation-review, /feel-pass, /build-playability-review, /game-qa, /game-ship, /game-debug, /game-retro, /game-codex, /game-docs, /game-visual-qa, /asset-review, /playtest, /careful, /guard, /unfreeze. Then ask the user if they also want to add gstack-game to the current project.
- ### Option B: Add to your repo so teammates get it
- > Add gstack-game to this project: run **`cp -Rf ~/.claude/skills/gstack-game .claude/skills/gstack-game && rm -rf .claude/skills/gstack-game/.git && cd .claude/skills/gstack-game && bun run build`** then add a "gstack-game" section to this project's CLAUDE.md listing all available skills.
- Real files get committed to your repo (not a submodule), so `git clone` just works. Everything lives inside `.claude/`. Nothing touches your PATH or runs in the background.
- ### Option C: Manual install
- ```bash
- # Install a single skill:
- cp -r /path/to/gstack-game/skills/game-review your-project/.claude/skills/
- # Or install everything:
- /path/to/gstack-game/bin/install.sh .
- ```
- ---
- ## See it work
- ```
- You: I have a game design PDF I'd like to review
- You: /game-import
- Claude: I found your PDF. Let me read it and check what's there.
- | Section | Status |
- | Core Loop | ✅ |
- | Systems Detail | ✅ |
- | Progression | ❌ |
- | Economy | ❌ |
- | Player Motivation | ❌ |
- 3/8 sections missing. Want me to ask about each gap?
- A) Probe the gaps — one question per missing section
- B) Just write the GDD — mark gaps as ❌
- C) Focus on critical gaps only
- You: A
- Claude: Your GDD doesn't define how players come back.
- What pulls a player back for a second session?
- A) New content unlocks
- B) Unfinished business — "my character is still growing"
- C) Daily rewards / rituals
- D) I don't know yet — mark as TBD
- → [walks through each gap with one focused question]
- → Writes docs/gdd.md (116 lines, 5/8 sections)
- → "Ready for /game-review"
- You: /game-review
- Claude: Mode: A (Mobile/Casual) — weights adjusted
- Section 1 — Core Loop: 7/10
- "The 4-phase loop is clear but meso-loop is undefined."
- A) Continue to Progression B) Dig deeper C) Fast-forward
- → [walks through each section, one issue at a time]
- GDD Health Score: 5.8/10
- STATUS: DONE_WITH_CONCERNS
- You: /player-experience
- Claude: Who should I pretend to be?
- A) Casual Newcomer (3-min attention)
- B) Interested Returner (Day 2-3)
- → [recommends A based on GDD state]
- Phase 1 — First Contact (0-30s):
- 0:00 — I see a loading screen. Emotion: Neutral
- 0:12 — First tap. Emotion: Engaged ✅ Aha moment
- 1:30 — Waiting for next wave. Emotion: Bored 🔴
- "Would this persona return? My assessment: unlikely — 1:30 dead time."
- You: I want to brainstorm a new concept instead
- You: /game-ideation
- Claude: First — what stage are you at? Do you have a prototype, or just the concept?
- → Level 1: one-line pitch. Let's go deeper.
- Fantasy: What should the player FEEL? "Run a dungeon" is a mechanic,
- not a feeling. Is it "the satisfaction of watching your empire grow
- from nothing"? Or "the evil joy of watching adventurers fall into
- your traps"?
- → You pick: evil joy + strategic satisfaction
- Core Loop: Describe it in one sentence:
- "Place traps → watch adventurers trigger them → collect loot → upgrade → repeat"
- Is the 30-second micro-loop fun on its own? Or only fun because of upgrades?
- → Forcing Q: If you remove all upgrades, is "place → trigger" still fun?
- Twist: Name the 3 most similar games. How are you different?
- Is the difference in mechanics (hard to copy) or art (easy to copy)?
- → Outputs Concept One-Pager
- You: /game-review
- Claude: [Reads Concept One-Pager]
- Mode A (Mobile/Casual) — weights adjusted
- Core Loop: 7/10 — micro-loop clear but meso-loop undefined
- Progression: 5/10 — D7 retention hook not defined (-3)
- Economy: 4/10 — faucets only, no sinks (inflation risk)
- GDD Health Score: 5.8/10
- STATUS: DONE_WITH_CONCERNS
- [Walks through each issue one at a time via AskUserQuestion]
- You: /player-experience
- Claude: Persona: Casual Newcomer (3-min commute session)
- 0:00 Open app Curious
- 0:05 Loading done Tutorial text ⚠️ Wall of text, 3+ lines = skipped
- 0:12 First trap placed Engaged ✅ Aha moment
- 0:30 First adventurers Excited ✅ Seeing results
- 1:30 Waiting for wave Bored 🔴 60s dead time, churn risk
- EMOTION ARC: Curious → Excited → Bored
- Issue: 1:30 dead time will kill D1 retention
- You: /gameplay-implementation-review
- Claude: Diff: 847 LOC, 12 files
- [Pass 1 — Critical]
- 🔴 DungeonManager.Update() — new List<>() every frame, GC pressure
- 🔴 TrapData serialization has no version field — save upgrades will break
- ⚠️ Damage calc on client side — cheat vector in multiplayer
- [Pass 2 — Informational]
- Hardcoded magic number: trap_damage = 15
- AUTO-FIXED: 2 issues
- ASK: 3 issues (one at a time)
- ```
- Six commands. Concept to code review. Each step's output feeds the next.
- ---
- ## The sprint
- gstack-game is a process, not a collection of tools. Skills are ordered the way a game dev sprint runs:
- **Spark → Think → Plan → Review → Slice → Handoff → Build → Feel → Playability → Test → Ship → Reflect**
- Each skill feeds the next. `/spark-lens` can leave an optional spark trace before structure. `/game-import` converts your PDF into a GDD that `/game-review` reads. `/game-review` flags risks that `/prototype-slice-plan` uses to decide what to build first. `/implementation-handoff` packages the slice for coding. `/feel-pass` checks if the build feels alive. `/build-playability-review` asks "is this worth playing?" before QA. Structured outputs are saved to `~/.gstack/projects/` so downstream skills find them automatically — even across sessions.
- | Skill | Your specialist | What they do |
- |-------|----------------|--------------|
- | `/triage` | **Project Navigator** | Detects project state (BLANK/IDEA/DOCUMENTED/REVIEWED/BUILDING/SHIPPING), routes to the right skill. Run this first if unsure. |
- | `/spark-lens` | **Creative Spark Companion** | Protects fragile early ideas, turns moods/images/mechanics into creative riffs, finds the emotional imprint and optional soul slice without critique or scoring |
- | `/game-import` | **Document Specialist** | Convert any format (PDF, Notion, verbal description) into standardized `docs/gdd.md`. Audit completeness against 8-section standard. Pipeline gateway. |
- | `/game-ideation` | **Game Design Mentor** | Structure concepts with Fantasy/Loop/Twist, challenge assumptions with 6 forcing questions, plan next validation step with Iceberg framework |
- | `/game-direction` | **Producer / Creative Director** | Challenge "why build this game?", 10 cognitive patterns for strategic review, scope decisions (ADD/KEEP/DEFER/CUT) |
- | `/game-review` | **Senior Game Designer** | GDD review: Core Loop, Progression, Economy, Motivation, Risk. Quantified GDD Health Score |
- | `/game-eng-review` | **Technical Director** | Engine choice, rendering pipeline, networking architecture, asset pipeline, platform adaptation |
- | `/balance-review` | **Economy Designer** | Difficulty curves (Flow State), Sink/Faucet economy, Gini coefficient, pity systems, reward psychology |
- | `/player-experience` | **UX Researcher** | First-person player walkthrough, 7 personas (4D: profile/stance/behavior/influence), emotion model, repeat play simulation |
- | `/game-ux-review` | **UI/UX Designer** | HUD readability, menu flow, shop UI, tutorial, controller/touch adaptation, accessibility |
- | `/plan-design-review` | **Senior Product Designer** | Pre-implementation plan review: 7-pass Fix-to-10 (info architecture, interaction states, player journey, AI slop, design system, input/a11y, decisions). Builds DESIGN.md from scratch. Codex + Subagent parallel review |
- | `/pitch-review` | **Investor / Publisher Lens** | Market positioning, differentiation, feasibility, business case, Iceberg validation level |
- | `/prototype-slice-plan` | **Production Strategist** | Decide what to build FIRST: which slice, what hypothesis to test, what to fake, what failure looks like |
- | `/implementation-handoff` | **Design Translator** | Convert design intent into a build package: gameplay requirements, acceptance criteria, MUST/SHOULD/COULD priority |
- | `/gameplay-implementation-review` | **Senior Game Programmer** | Three-pass review: Pass 0 (design intent survival) + Pass 1 (frame budget, memory, sync) + Pass 2 (code quality). Evolved from /game-code-review |
- | `/feel-pass` | **Game Feel Doctor** | Diagnose why a mechanic feels dead: responsiveness, impact, rhythm, clarity, payoff. 7-dimension /14 scoring. The most game-specific skill. |
- | `/build-playability-review` | **Playability Judge** | "Is this worth playing?" — loop closure, session viability, retention signal, peak moments. 6-dimension /12 scoring |
- | `/game-qa` | **QA Lead** | 8-dimension testing: functional, visual, performance, audio, input, compatibility, localization, progression. Quantified Health Score |
- | `/game-ship` | **Release Engineer** | Build → Test → Changelog → PR → platform submission (Steam/App Store/Google Play/Web) |
- | `/game-debug` | **Debug Specialist** | 3-strike hypothesis testing + root cause analysis. No guessing. |
- | `/game-retro` | **Scrum Master** | Delivery rate, bug density trends, velocity, milestone health. Max 3 action items. |
- | `/game-codex` | **Chaos Engineer** | Independent-context adversarial review. Finds exploits, desync, save corruption. |
- | `/game-docs` | **Technical Writer** | Player-friendly patch notes + internal changelog + doc sweep |
- | `/game-visual-qa` | **Art QA** | Style consistency, UI alignment, animation quality, screen adaptation |
- | `/asset-review` | **Technical Artist** | Naming conventions, format specs, performance budget, style consistency |
- | `/playtest` | **UX Researcher** | Test plans, observation metrics, interview questions, data analysis framework |
- ### Safety tools
- | Skill | What it does |
- |-------|-------------|
- | `/careful` | Warns before destructive commands (rm -rf, force push, DROP TABLE) |
- | `/guard` | `/careful` + restricts file edits to a specific directory |
- | `/unfreeze` | Removes `/guard` scope restriction |
- ---
- ## Relationship to gstack
- | | gstack | gstack-game |
- |---|--------|-------------|
- | **Domain** | Web / SaaS products | Game development |
- | **Vocabulary** | user, feature, API, MRR, churn | player, mechanic, core loop, retention, ARPDAU |
- | **Dependency** | Standalone | Standalone (no gstack required) |
- **Borrowed from gstack:**
- - Template Engine (SKILL.md.tmpl → gen-skill-docs.ts → SKILL.md)
- - Preamble Injection (shared session tracking, telemetry, AskUserQuestion format across all skills)
- - 6 review methodology principles (classify-before-judge, explicit scoring, AUTO/ASK/ESCALATE, structured questions, multi-dim cross-check, anti-sycophancy)
- - Completion Protocol (DONE / DONE_WITH_CONCERNS / BLOCKED / NEEDS_CONTEXT)
- **Fully rewritten:**
- - All review dimensions, scoring criteria, and forcing questions → game domain
- - 6 game-specific skills that gstack doesn't have
- - Game theory integration (MDA Framework, SDT, Flow State, Nested Loop Model, Sink/Faucet economy)
- ---
- ## Quality assessment
- Each skill's percentage indicates **domain judgment completeness**, not code quality.
- | Score | Meaning |
- |-------|---------|
- | 70-80% | Full structure + domain theory + quantitative scoring. **Ready to use, expert fine-tuning only** |
- | 55-65% | Full structure + game vocabulary + AUTO/ASK/ESCALATE. **Usable but lacks domain depth** |
- | 35-40% | Structural skeleton + basic flow. **Needs domain expert input** |
- The engineering backbone matches gstack quality. What's missing is **game industry domain judgment** — scoring weights, benchmark numbers, engine-specific review points. These need calibration from experienced game professionals. See `docs/domain-judgment-gaps.md` for the full checklist.
- ### Theory sources
- | Source | What we took |
- |--------|-------------|
- | **[gstack](https://github.com/garrytan/gstack)** | Engineering architecture + 6 review methodology principles |
- | **Claude-Code-Game-Studios** | MDA/SDT/Flow theory, Nested Loop, Pillar methodology, economy frameworks |
- | **guardian** (PlayerSimulatorAgent) | Player simulation prompts, Iceberg validation framework, Fantasy/Loop/Twist |
- | **hakoniwa** | 4D persona model (stance/MBTI-behavior/influence), ReACT evidence methodology, community reception prediction, event injection testing |
- ---
- ## Development
- ```bash
- bun run build # Generate all SKILL.md from templates
- bun run gen:skill-docs:check # Check for drift (CI use)
- bun test # Run 24 validation tests
- ```
- ### Adding a new skill
- 1. Create `skills/my-skill/SKILL.md.tmpl`
- 2. Use `{{PREAMBLE}}` for shared behavior injection
- 3. Use `{{SKILL_NAME}}` for auto-filled skill name
- 4. Run `bun run build`
- ### File structure
- ```
- gstack-game/
- ├── CLAUDE.md ← Full technical docs
- ├── README.md ← This file
- ├── README.zh-TW.md ← 繁體中文版
- ├── ETHOS.md ← Game dev philosophy (Boil the Lake, game edition)
- ├── CHANGELOG.md ← Version history
- ├── VERSION ← 0.5.0
- ├── package.json ← Build scripts
- ├── bin/ ← 7 utilities (config, diff-scope, telemetry...)
- ├── scripts/gen-skill-docs.ts ← Template engine
- ├── skills/ ← 29 skills + shared preamble (many with references/)
- ├── test/ ← Tier 1 validation tests
- └── docs/ ← Domain judgment gaps, source assessment, dev guide
- ```
- ---
- ## Privacy & Telemetry
- gstack-game includes **opt-in** usage telemetry, disabled by default.
- - **Default is off.** Nothing is sent anywhere unless you explicitly opt in.
- - **What's recorded (if enabled):** skill name, duration, success/fail. That's it.
- - **Never recorded:** code, file paths, repo names, prompt content.
- - **Change anytime:** `gstack-config set telemetry off`
- - **Local only:** All data stored in `~/.gstack/analytics/skill-usage.jsonl`, never leaves your machine.
- ---
- ## Troubleshooting
- **Skill not showing up?** Make sure SKILL.md files are generated: `cd .claude/skills/gstack-game && bun run build`
- **Template drift?** `bun run gen:skill-docs:check` — if it reports DRIFT, run `bun run build` to regenerate.
- **Windows users:** Run under Git Bash or WSL. The shell scripts in bin/ require bash.
- **Claude doesn't see the skills?** Add this to your project's CLAUDE.md:
- ```
- ## gstack-game
- Available skills: /triage, /spark-lens, /game-import, /game-ideation, /game-direction,
- /game-review, /game-eng-review, /balance-review, /player-experience,
- /game-ux-review, /plan-design-review, /pitch-review, /prototype-slice-plan, /implementation-handoff,
- /gameplay-implementation-review, /feel-pass, /build-playability-review,
- /game-qa, /game-ship, /game-debug, /game-retro, /game-codex, /game-docs,
- /game-visual-qa, /asset-review, /playtest, /careful, /guard, /unfreeze.
- ```
- ---
- ## Docs
- | Doc | What it covers |
- |-----|---------------|
- | [Builder Ethos](ETHOS.md) | Game dev philosophy: Boil the Lake, Search Before Building, Player Time is Sacred |
- | [Contributing](CONTRIBUTING.md) | Three ways to contribute (5min Issue / 30min markdown / advanced template) |
- | [Domain Judgment Gaps](docs/domain-judgment-gaps.md) | Expert review checklist — what needs calibration |
- | [Development Guide](docs/DEVELOPMENT.md) | Skill map, migration guide, architecture overview |
- | [Changelog](CHANGELOG.md) | Version history |
- | [CLAUDE.md](CLAUDE.md) | Developer handoff for AI agents working on this repo |
- ---
- ## License
- MIT
.claude/settings.json · Bundled script · 47 lines · 1 finding marked
- {
- "defaultMode": "acceptEdits",
- "permissions": {
- "allow": [
- "Bash(git *)",
- "Bash(gh *)",
- "Bash(bun *)",
- "Bash(node *)",
- "Bash(npm *)",
- "Bash(npx *)",
- "Bash(ls *)",
- "Bash(cat *)",
- "Bash(mkdir *)",
- "Bash(cp *)",
- "Bash(mv *)",
- "Bash(find *)",
- "Bash(head *)",
- "Bash(tail *)",
- "Bash(grep *)",
- "Bash(diff *)",
- "Bash(wc *)",
- "Bash(sort *)",
- "Bash(uniq *)",
- "Bash(sed *)",
- "Bash(awk *)",
- "Bash(cd *)",
- "Bash(echo *)",
- "Bash(pwd)",
- "Bash(which *)",
- "Bash(touch *)",
- "Bash(chmod *)",
- "Bash(xargs *)",
- "Bash(for *)",
- "Bash(rm *)",
- "Edit",
- "Read",
- "Write"
- ],
- "deny": [
- "Bash(rm -rf /)",
- Dismissed after review: Runs `rm -rf` against a root or home-directory path
- Line 40 on GitHub
- "Bash(sudo:*)",
- "Read(.env)",
- "Read(.env.*)",
- "Read(~/.ssh/**)"
- ]
- }
- }
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.