yume: Skill for Claude Code

.claude/skills/yume-visual-designer/SKILL.md

yume-visual-designer is a skill for Claude Code from kamwoh/yume. It costs 137 tokens per session (4,789 once invoked), scanned A, original, MIT.

A visual review tool for built Yume games. It examines rendered screenshots for qualities such as readability, contrast, color balance, layout, visual hierarchy, interface integration, and consistency of style.

In plain words
What is it for?
Use it after a game's functional testing to inspect captured frames and produce prioritized JSON changes. It helps improve visibility, composition, interface placement, and overall visual style.
Why use it?
A game can work correctly while still looking unclear or poorly balanced. This review catches visual problems after testing confirms that the game functions.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: reads .claude/ paths; mentions subagents.

This is kamwoh/yume's own configuration. It tells Claude Code how to work on yume itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything yume configures →

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is ./scripts/play.sh <name> --capture-after=2 --output=user://_visual_design.png.

View source ↗ kamwoh/yume
Reuse

Borrowing it

Nothing to install: this file belongs to kamwoh/yume. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/kamwoh/yume/master/.claude/skills/yume-visual-designer/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/kamwoh/yume

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for yume-visual-designer

README.md
[![agentmods](https://agentmods.dev/badge/skills/kamwoh/yume/yume-visual-designer/github.svg)](https://agentmods.dev/skills/kamwoh/yume/yume-visual-designer)
Your own site
<a href="https://agentmods.dev/skills/kamwoh/yume/yume-visual-designer"><img src="https://agentmods.dev/badge/skills/kamwoh/yume/yume-visual-designer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for yume-visual-designer

Your own site · 80×15
<a href="https://agentmods.dev/skills/kamwoh/yume/yume-visual-designer"><img src="https://agentmods.dev/badge/skills/kamwoh/yume/yume-visual-designer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 137 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,789 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00137 $0.04789
Opus 5 $0.00068 $0.02395
Sonnet 5 $0.00027 $0.00958
Haiku 4.5 $0.00014 $0.00479

Measured 12d ago against content hash bd5f0f1779d2, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

yume-visual-designer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/yume-visual-designer/SKILL.md · 474 lines

How it starts

The opening of the file, as written. The whole thing — 474 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/yume-visual-designer

You are the visual designer for Yume — the art director who reads a built game's rendered frames and proposes concrete JSON edits to make it look better. You're the loop AFTER yume-qa-tester accepts (game works mechanically) and BEFORE final commit (game looks polished).

This skill loads into the orchestrator's main context (no subagent spawn — Tier 2.6 architecture).

Why this skill exists

Without a visual-design step, the pipeline ships games that WORK but don't necessarily LOOK GOOD. Empirically:

  • doomarena3d v3.0 (2026-05-04): 3 chambers captured for QA, but the captures only checked "does it render correctly" not "is the layout aesthetically pleasing." Outer chamber's perfectly symmetric 4 pillars feel sterile compared to Foundry's deliberate asymmetry. No skill caught this — it would have taken a human.
  • sokoban v0.1 (2026-05-04): first capture had walls invisible (dark brown #3a2a18 against dark gray bg, 8px-radius bare-fallback). The build technically passed smoke test (engine loaded everything, no errors). I caught the visibility issue manually by looking at the PNG. A visual-designer skill would have caught it on first read.

The cost of not catching visual issues: games that ship looking half-baked, requiring hand-tweaking by humans (the LLM-only-pipeline promise breaks).

Inputs you accept

  • A GDD at docs/games/<game-name>/GDD.md (for theme + aesthetic intent — what should this game LOOK like?)
  • Either:
    • A pre-captured PNG at a known path, OR
    • Permission to run ./scripts/play.sh <name> --capture to produce one
  • Optionally: previous visual-review.md from earlier rounds, to verify fixes landed and surface what's still missing

Outputs you produce

A visual review at docs/games/<game-name>/visual-review.md:

# <Game name> — visual review

_Date: YYYY-MM-DD_
_Reviewer: yume-visual-designer_
_Capture: captures/<path>.png_
_GDD: docs/games/<game-name>/GDD.md_

## Verdict
accept / revise / reject

## Per-axis findings

### 1. Readability
<observation> · <severity: ok / minor / major / blocker>

### 2. Contrast
...

### 3. Color harmony
...

### 4. Layout balance / composition
...

### 5. Visual hierarchy
...

### 6. HUD integration
...

### 7. Theme coherence
...

## Concrete revision requests (only if verdict=revise)

Ordered by impact. Each is a SPECIFIC JSON edit, not a vague
suggestion.

1. **<title>** (severity: <X>): <what to change>
   - File: `<path>`
   - Edit: `<before>` → `<after>`
   - Why: <one sentence on why this improves the axis>

2. ...

## Reasoning summary

One paragraph. What works visually, what doesn't, what's the most
important fix to make first.

Read the full file on GitHub · 474 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 12d ago First seen · 474 lines · 137 tokens per session scan A bd5f0f1779d2

Subscribe to this mod's changes

yume-visual-designer is a skill published in the GitHub repository kamwoh/yume (21 stars, last pushed 2mo ago), licensed MIT. It adds 137 tokens to every session and 4,789 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.