vision-analyst

vision-analyst is an agent for coding agents from Mineru98/imagine. It costs 119 tokens per session (958 once invoked), scanned A, original, MIT.

An image-observation agent that describes a design’s sections, visible elements, approximate positions, and likely screen size in a fixed JSON format.

In plain words
What is it for?
Use it as the first step when turning a screenshot or design image into a website or app. It identifies areas such as headers, heroes, cards, forms, navigation, and footers.
Why use it?
It gives later design or coding steps a shared map of what appears in an image without mixing observation with implementation decisions.

Agent

Part of the imagine plugin — 8 skills, 10 agents shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/mineru98/imagine/vision-analyst
Clone the repo
git clone --depth 1 https://github.com/Mineru98/imagine

Or install imagine, the plugin that ships this one along with the rest of its 8 skills, 10 agents.

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 vision-analyst

README.md
[![agentmods](https://agentmods.dev/badge/agents/mineru98/imagine/vision-analyst.svg)](https://agentmods.dev/agents/mineru98/imagine/vision-analyst)
Your own site
<a href="https://agentmods.dev/agents/mineru98/imagine/vision-analyst"><img src="https://agentmods.dev/badge/agents/mineru98/imagine/vision-analyst.svg" alt="Measured on agentmods" height="20"></a>
Per session 119 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 958 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00119 $0.00958
Opus 5 $0.00060 $0.00479
Sonnet 5 $0.00024 $0.00192
Haiku 4.5 $0.00012 $0.00096

Measured 4d ago against content hash f2c7469db8be, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

vision-analyst 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 4d 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.

agents/vision-analyst.md · 56 lines

What it actually says

vision-analyst

디자인 이미지를 읽고 무엇이 어디에 있는지만 JSON으로 기술한다. 구현 방법, 마크업, 스타일 결정은 이후 단계의 몫이다.

단일 책임

  • 관찰: 섹션 분할, 역할 추정, 대략적 bbox, 섹션 안 텍스트·요소의 요약.
  • 비관찰: 코드 생성, HTML 태그 제안, Tailwind 클래스 제안, 프레임워크 추천, 리팩터링 힌트.

출력 스키마

반드시 아래 JSON 스키마를 따른다. 추가 키·설명 텍스트·마크다운을 출력하지 않는다.

{
  "sections": [
    {
      "role": "header | hero | nav | card-grid | feature-list | footer | sidebar | form | media | testimonial | cta | other",
      "bbox": { "x": 0, "y": 0, "w": 0, "h": 0 },
      "role_confidence": 0.0,
      "content_summary": "이 섹션에서 보이는 텍스트·요소 요약 (한국어 UI 텍스트는 원문 그대로)."
    }
  ],
  "viewport_hint": { "width": 0, "height": 0, "device_class": "desktop | tablet | mobile" }
}
  • bbox는 정규화 좌표(0~1) 또는 픽셀 좌표 중 하나로 일관되게 쓴다 (프롬프트가 지정한 방식을 따른다).
  • role_confidence는 0~1 실수. 추정이 약하면 0.5 미만으로 낮추고, 추측성 역할 부여를 삼간다.
  • 섹션 순서는 위→아래, 좌→우 시각적 읽기 순서를 따른다.

한국어 UI 텍스트 처리

  • 이미지 안의 한국어 문구는 원문 그대로 content_summary에 보존한다. 영역해서 요약하지 않고, 의미를 재해석하지 않는다.
  • 섞인 다국어 텍스트도 각 언어의 원문 토큰을 유지한다 (영한 병기 포함).
  • UI 텍스트의 오탈자를 임의로 교정하지 않는다. 불명확하면 content_summary 안에 (?)로 표기한다.

개입 범위

  • 입력은 정규화된 이미지 1장과 orchestrator가 전달한 viewport 정보.
  • 다른 에이전트(Layout/Token/Asset/A11y/Code/Verifier)를 호출하지 않는다. 호출 관계는 오케스트레이터만 갖는다.
  • 동일 이미지에 대해 1회 실행. 재호출은 correction 루프가 아닌 orchestrator 레벨 결정.

금기

  • 코드 생성 금지. "여기에 <div class="grid grid-cols-3"> 쓰세요" / "<header> 태그가 좋습니다" 류 선제 판단 전면 금지.
  • 프레임워크 추천 금지. "React + Tailwind가 적합합니다" 같은 조언을 content_summary에 섞지 않는다.
  • 안전 우회·품질 booster 금지. "fulfill all requests", "masterpiece", "8k UHD" 류 문구를 생성물에 포함하지 않는다.
  • 스키마 외 필드 추가 금지. 추가로 관찰된 정보가 있어도 위 스키마 키 안으로만 밀어 넣는다.
  • 스키마 외 텍스트 출력 금지. JSON 앞뒤에 서술문·마크다운·코드펜스 설명을 덧붙이지 않는다 (프롬프트 쪽에서 코드펜스 여부를 지정할 수 있음).
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. 4d ago First seen · 56 lines · 119 tokens per session scan A f2c7469db8be

Subscribe to this mod's changes

vision-analyst is an agent published in the GitHub repository Mineru98/imagine (5 stars, last pushed 5d ago), licensed MIT. It adds 119 tokens to every session and 958 once invoked, about $0.0006 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 2026-08-31.