Skip to content
Agentmart
Submit a tool

game-design

Visual target generation, risk-first game decomposition, and verification criteria.

Skill by dragoscirjan

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/dragoscirjan-game-design

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Visual target generation, risk-first game decomposition, and verification criteria. Load at pipeline start before architecture.

Use cases: design, data

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/game-design

What's inside

Skill name
game-design
Description
Visual target generation, risk-first game decomposition, and verification criteria. Load at pipeline start before architecture.
License
MIT
Instructions
152 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Game Design — Visual Target & Decomposition

## Overview

Two-phase design process that precedes all implementation:

1. **Visual Target** — generate a reference screenshot that anchors art direction
2. **Decomposition** — analyze risks, define verification criteria, produce a game plan spec (via `spec-create`)

---

## Phase 1: Visual Target

Generate a reference image of what the finished game looks like. Anchors art direction for scaffold, asset planner, and task agents.

### Generate

Use the `godot-asset-gen` tool:

```
godot-asset-gen image --model gemini --prompt "{prompt}" --size 1K --aspect-ratio 16:9 -o reference.png
```

### Prompt Rules

The reference must look like an **in-game screenshot**, not concept art. Every distinct object visible becomes an asset requirement downstream.

- **Enumerate every game object** — player, each enemy type, obstacles, collectibles, projectiles, platforms, props. Name each with position and approximate size. Objects absent from the reference get forgotten downstream.
- **Reflect real technical constraints** — tiling backgrounds should prompt tiling-friendly composition. Separate sprite layers should show distinct objects, not composited photorealism.
- **Do NOT prompt downgraded quality** ("lowpoly", "pixel art", "retro") — produces worse output. Prompt clean, sharp rendering with the actual composition needed.
- **Focus on the most important gameplay moment** — the frame showing spatial layout, core mechanic, and the camera perspective the player sees most.
- **Exclude what you will NOT build** — volumetric lighting, motion blur, depth of field, atmospheric fog, complex reflections, lens flares, detailed cast shadows. They create asset requirements nobody can fulfill.
- **Show HUD/UI elements** — health bar, score counter, minimap, inventory slots with screen positions. These are implementation requirements too.

### Prompt Template

```
Screenshot of a {2D/3D} video game. {Camera: angle, distance, perspective}.
Game objects: {player — appearance, position, size vs screen}. {enemies/NPCs — each type, position}. {obstacles}. {collectibles/pickups}. {projectiles if any}.
Environment: {background layers — sky, distant, mid}. {playfield surface — material, tiling}. {foreground elements}. {boundaries/edges}.
HUD: {each UI element — type and screen position}.
{Art style, color palette}. Clean sharp digital rendering, game engine output.
```

### Output

- `reference.png` — 1K 16:9 image
- Write art direction into a draft (via `draft-create`):

```markdown
# Assets

**Art direction:** <the art style description>
```

---

## Phase 2: Decomposition

Analyze the game for implementation risks and define verification criteria. Output is a game plan spec (created via `spec-create` with `type=task`).

### Workflow

1. **Read `reference.png`** — camera angle, scene complexity, entity count, environment scope.
2. **Read the game description** — core technical requirements.
3. **Scan for risks** — identify features needing isolation (see taxonomy).
4. **Define verification criteria** — risk-specific, general, and final.
5. **Write the game plan spec** (via `spec-create` with `type=task`, `author=game-director`).

### Risk Taxonomy

#### ISOLATE (fail unpredictably, ambiguous errors when mixed)

- **Procedural generation** — terrain, levels, meshes, dungeon layouts
- **Procedural animation** — runtime bone manipulation, IK, ragdoll blending
- **Sprite/character animations** — multi-direction movement, state transitions (almost always fail first pass)
- **Complex vehicle physics** — wheel colliders, suspension, drifting, motorcycle balance
- **Custom shaders** — water surfaces, portals, screen-space effects, dissolve/distortion
- **Runtime geometry** — destructible environments, CSG operations, mesh deformation
- **Dynamic navigation** — pathfinding adapting to runtime obstacles, crowd simulation, flocking
- **Complex camera systems** — third-person with collision avoidance, cinematic rail transitions, split-screen

#### NEVER ISOLATE (Godot handles well)

CharacterBody movement, collision/triggers, TileMap/GridMap, NavigationAgent on static navmesh, UI with Control nodes, spawning/timers/waves, camera follow, state machines, input handling.

### Verification Criteria

Each task gets a **Verify** field — what to check after implementation.

**Risk tasks** — target the exact failure mode:

- Animations: "every direction plays correct frames, transitions smooth, no pose snapping"
- Procedural gen: "output covers expected area, no gaps, no overlaps, no degenerate geometry"

**Main build** — combine cross-cutting with game-specific checks:

- Movement direction matches player input
- Animation direction matches movement direction
- Player input → character response feels correct
- Physics objects respond to gravity/collision
- UI readable, no overflow or overlap
- No missing textures (magenta/checkerboard)
- Game-specific checks (e.g., "enemies path around towers," "score increments on pickup")
- `reference.png` consistency
- Presentation video as final deliverable

### Game Plan Format

```markdown
# Game Plan: {Name}

## Game Description

{Original description, verbatim.}

## Risk Tasks

{Omit entirely if no risks identified.}

From dragoscirjan/opencode-config, MIT.

More from dragoscirjan/opencode-config

Signals

Stars
0
Forks
0
Open issues
3
Last commit (this folder)
May 28, 2026 (5 months ago)
Last release
Jun 2, 2026 (4 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from dragoscirjan

Cite this page

game-design skill. Agent Mart. https://agentmart.dev/skills/dragoscirjan-game-design (last updated 2026-10-24)