| name | game-research |
| description | Structured game development research with solo-dev feasibility filtering. Use when investigating engines, genres, art styles, tools, or any game-dev decision. Wraps the research skill with game-specific framing. Outputs feed into decision-journal. |
| zones | {"knowledge":30,"process":35,"constraint":15,"wisdom":20} |
game-research
Research for someone who is learning to build, not someone who already knows how.
Before Starting
- Invoke the
research skill. This skill extends it, not replaces it. The research skill's stance, evidence hierarchy, and citation requirements all apply.
- Invoke the
decision-journal skill. Know the output format before gathering.
The Solo Dev Filter
Every research direction must pass through this question:
Can one person, who is still learning, achieve this within a reasonable timeline?
This filter eliminates:
- AAA production pipelines requiring teams of 50+
- Technologies with 2+ year learning curves before productive use
- Art styles requiring professional training the user doesn't have
- Scope that assumes full-time development when this is a side project
When presenting findings, always include a feasibility assessment for a solo learner.
Research Domains
Domain Dependency Map
Genre & Scope
├── Engine (what can run it?)
│ └── Language (what does the engine use?)
│ └── Learning Path (how do I learn it?)
├── Art Style (what fits the genre?)
│ └── Art Tools (what produces this style?)
└── Sound & Music (what does the genre need?)
Research in dependency order. Don't research art tools before knowing the art style. Don't research languages before knowing the engine.
Domain Reference
| Domain | Key Questions | Solo Dev Traps |
|---|
| Genre & Scope | What's achievable? What sells on Steam? What do you enjoy? | Scope creep. "Just one more feature" syndrome. Comparing to AAA. |
| Engine | Learning curve? Community? Export to Steam? Language support? | Choosing based on capability ceiling, not learning floor. |
| Language | Learning resources? boot.dev courses? Transferable skills? | Picking the "best" language instead of the most learnable. |
| Art Style | Achievable solo? Consistent at scale? Asset pipeline? | Pixel art "looks easy" but consistency is hard. 3D has a huge tool learning curve. |
| Art Tools | Cost? Learning curve? Export formats? Community? | Tool hopping. Buying expensive tools before committing to a style. |
| Sound & Music | Create vs license vs AI? Legal implications? | Underestimating audio's importance, then scrambling at the end. |
| Game Design | GDD depth? Prototyping approach? Scope control? | Over-designing before prototyping. GDD as procrastination. |
| SDLC | Version control? CI/CD? Testing? | Over-engineering the pipeline before there's code to run through it. |
| Steam | Steamworks setup? Store page timing? Wishlists? | Marketing too late. No store page until launch. |
| Learning | boot.dev path? Supplementary? Practice projects? | Tutorial hell. Learning without building. |
| Community | When to start? Discord? Devlog? | Starting too early (nothing to show) or too late (no audience). |
Research Process
1. Frame the question
What decision will this research inform? Be specific.
- Good: "Which game engine has the best learning curve for a beginner targeting 2D games on Steam?"
- Bad: "What game engine should I use?"
2. Identify the domain and dependencies
Check the dependency map. Are prerequisite decisions already made? If not, research those first.
Check the decision-journal — has this already been decided? If yes, check the revisit trigger before proceeding.
3. Define research directions
For each domain, consider:
- Technical capability — Can it do what I need?
- Learning curve — How long to become productive?
- Community & resources — Tutorials, forums, Stack Overflow presence?
- Solo dev track record — Have solo devs shipped with this? Examples?
- Cost — Free? One-time? Subscription? Revenue share?
- Longevity — Is this actively maintained? Growing or dying?
4. Conduct research
Follow the research skill's process:
- Launch parallel subagents for different directions
- Include one subagent for "directions I haven't thought of"
- Apply the solo dev filter to all findings
5. Synthesize
Present findings using the research skill's output format. Add:
- Solo dev feasibility: explicit rating (high/medium/low) with justification
- Learning investment: estimated time to become productive (not to mastery)
- Decision recommendation: what the evidence points toward (synthesis, not prescription)
6. Record the decision
After the user decides, invoke decision-journal to create an entry with:
- Alternatives considered (from research)
- Evidence links (from research output)
- Revisit trigger (from the findings)
- Confidence level (based on research depth)
Steam Market Context
When researching genres, include Steam market data:
- What sells for solo devs? Look at indie success stories, not AAA.
- Saturated vs underserved genres — 2D platformers are saturated. Niche genres may have smaller but more passionate audiences.
- Price expectations by genre — An RPG can charge $20. A puzzle game struggles above $10.
- Wishlist-to-sale conversion — Industry benchmarks for indie games.
- Time to recoup — How many units at what price to cover costs?
Sources: SteamDB, Steam Spy (limited), GDC talks, indie postmortems, How to Market a Game (Chris Zukowski).
Constraints
- Apply the solo dev filter to every finding. Don't present options that require a team.
- Cite sources for every claim. The
research skill's evidence hierarchy applies.
- Present synthesis, not prescription. "The evidence suggests..." not "You should..."
- Check the decision-journal before researching. Don't re-research settled decisions unless the revisit trigger has been met.
- Account for learning curve, not just capability. "Can do X" means nothing if it takes 2 years to learn.
Deeper
- The
research skill — Core research methodology and subagent coordination
- The
decision-journal skill — Where decisions land after research
references/research-domains.md — Detailed questions for each domain
Research for a learner, not an expert. Feasibility before capability.