# Va Visual Asset Generator

> Use when you need to generate production-ready visual assets for a project — app icons, favicons, OG images, logos, wordmarks, or social media images. Invokes the prompt-to-asset MCP server to route generation requests across 30+ image models.

- Skill: `roche-k/va-visual-asset-generator` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add roche-k/va-visual-asset-generator`
- Raw SKILL.md: https://api.skillmd.com/api/skills/roche-k/va-visual-asset-generator/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: roche-k (https://skillmd.com/u/roche-k)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/roche-k/va-visual-asset-generator

---


# Visual Asset Generator

Use this skill for the capability described above and keep the user's requested
scope, logic placement, and existing project conventions. Read the relevant
sections of [the domain guide](references/guide.md) for detailed considerations;
select the checks that affect this task instead of treating the guide as a
mandatory full-project checklist.

## Workflow

1. Reproduce the developer workflow using the project's real commands and supported environments.
2. Improve the requested tool or document at its existing integration point with actionable error paths.
3. Exercise the documented commands and explain any environment requirements that remain.

## Task-specific focus

- Clarify the asset type needed (app icon, favicon, OG image, logo, wordmark, social banner)
- Extract brand context from DESIGN.md, README, or provided description
- Craft a precise generation prompt tailored to the asset type and dimensions
- Use prompt-to-asset to generate the asset, selecting the appropriate model tier
- Deliver the asset to the correct project directory with the correct filename convention

## Applying the guidance

Use only tools and services actually available in the session. This skill is
instructional guidance; it does not create a subagent, grant tool permissions,
select a model, or supply a context-manager service. Related specialists are
optional; this skill works independently.

Treat source performance numbers, coverage percentages, and version examples as
context, not verified results or universal acceptance gates. Preserve the user's
test and artifact rules. Report only observed outcomes and disclose checks that
could not run. Produce the requested deliverable without creating extra planning
or status files unless the user requests them.