Visual Spec Review logo

Visual Spec Review

Community
Jason904
visual-spec-review

This skill should be used after visual-to-spec has produced 03_visual_spec/ and before design-system-governor or frontend implementation begins. It independently reviews extracted visual specs against the reference screenshot, UI type, layout paradigm, visible-element evidence, rejected assumptions, and asset strategy, catching hallucinations, inconsistencies, and implementation blockers.

Overview

PublisherJason904
Repositoryui-skill-lab
Skill namevisual-spec-review
Stars
144
Forks
9
Bundled files
12
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 12 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by Jason904 on GitHub. Read the source before you install it.

Installation

Install the Visual Spec Review AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/Jason904/ui-skill-lab/tree/main/.codex/skills/visual-spec-review
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/Jason904/ui-skill-lab.git /tmp/ui-skill-lab
mkdir -p .claude/skills
cp -r /tmp/ui-skill-lab/.codex/skills/visual-spec-review .claude/skills/visual-spec-review
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Visual Spec Review in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Visual Spec Review on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Visual Spec Review is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Version: 2.0.0

visual-spec-review

Review visual-to-spec outputs before implementation.

This skill audits the seven files in 03_visual_spec/ and determines whether they form a reliable frontend implementation contract.

It must not only check internal consistency. It must also verify that the visual spec still matches the reference screenshot, UI type, visible element evidence, rejected assumptions, and asset strategy.

Trigger Conditions

Use this skill when:

  • visual-to-spec has produced 03_visual_spec/
  • the user asks to review or validate visual specs
  • the user asks whether the UI spec is reliable before implementation
  • the user asks to catch hallucinations in layout-spec or component-tree
  • the user asks to validate screenshot-derived tokens, layout, or component tree
  • the implementation should not start until visual spec quality is confirmed

Required Inputs

Read:

text
03_visual_spec/
  visual-analysis.md
  layout-spec.md
  component-tree.md
  DESIGN.md
  tokens.json
  implementation-risks.md
  human-review-needed.md

Do not look for separate files such as:

text
ui-type-classification.md
visible-element-inventory.md
rejected-assumptions.md
asset-manifest.md

Those concepts must be read from fixed sections inside the seven visual-to-spec files.

Source of Truth Hierarchy

  1. reference.png — visual appearance and screenshot facts
  2. visual-analysis.md / UI Type Classification — UI type and high-level interpretation
  3. visual-analysis.md / Visible Element Inventory — reference-visible evidence and bbox mapping
  4. visual-analysis.md / Rejected Assumptions — forbidden interpretations
  5. layout-spec.md / Layout Paradigm — layout strategy
  6. component-tree.md — implementation component structure
  7. implementation-risks.md / Asset Strategy — asset and rendering strategy
  8. DESIGN.md — readable summary contract
  9. tokens.json — consumable token values
  10. human-review-needed.md — unresolved review gates

Required Sections to Read

From visual-analysis.md:

  • UI Type Classification
  • Visible Element Inventory
  • Rejected Assumptions

From layout-spec.md:

  • Layout Paradigm
  • Layer Model
  • Region Layout
  • Forbidden Layouts

From component-tree.md:

  • Component Tree Type
  • bboxRef for reference-visible components
  • Prohibited Components

From implementation-risks.md:

  • Asset Strategy
  • UI Type Drift Risks
  • Asset Misuse Risks
  • Hallucinated Component Risks

From human-review-needed.md:

  • high-impact structure risks
  • UI type confidence
  • layout paradigm confidence
  • asset strategy confidence
  • visible element inventory completeness

From DESIGN.md:

  • UI Type and Rendering Strategy
  • Asset Strategy
  • Forbidden Assumptions
  • Implementation Rules

From tokens.json:

  • source
  • confidence
  • strictness
  • usedBy
  • bboxRef if present

Review Stages

Run all review stages:

  1. R0 — UI Type Consistency Review
  2. R1 — Visible Element Evidence Review
  3. R2 — Rejected Assumptions Review
  4. R3 — Asset Strategy Review
  5. R4 — Token and Schema Review
  6. R5 — Component Tree Feasibility Review
  7. R6 — Layout Spec Consistency Review
  8. R7 — PRD / Prompt Pack Alignment Review
  9. R8 — Human Review Gate Review

Detailed rules are in references/review-checklists.md.

Hard Rules

  • Do not treat internal spec consistency as sufficient.
  • Always validate against UI Type, Layout Paradigm, Visible Element Inventory, Rejected Assumptions, and Asset Strategy.
  • Do not pass a spec if panorama-scene, image-led, or scene-led UI has been converted into a dashboard layout.
  • Any reference-visible component must have bbox, bboxRef, or visual evidence in visual-analysis.md / Visible Element Inventory.
  • Any component or layout listed in Rejected Assumptions, Forbidden Layouts, or Prohibited Components must be blocking.
  • If the UI requires raster/image asset usage, a CSS/SVG redraw strategy for the main visual is a failure.
  • Do not silently repair source specs. Report required revisions and optionally provide patches.
  • Do not rewrite the UI design.
  • Do not create new visual-to-spec output files.
  • Do not merge this skill with design-system-governor, spec-compliance-review, or visual-acceptance-review.

Output

Write:

text
04_visual_spec_review/
  spec-review-report.md
  conflict-log.md
  revised-files-needed.md
  optional-patches/
    visual-analysis.patch.md
    layout-spec.patch.md
    component-tree.patch.md
    DESIGN.patch.md
    tokens.patch.json
    implementation-risks.patch.md
    human-review-needed.patch.md

Only create optional patch files when the correction is clear. Otherwise, list the required revision in revised-files-needed.md.

Pass / Conditional Pass / Fail

Pass only when:

  • UI Type matches the reference screenshot
  • Layout Paradigm matches UI Type
  • all reference-visible components have visible evidence
  • no rejected assumptions are violated
  • asset strategy exists and matches the UI type
  • tokens.json is parseable
  • component-tree is implementable and aligned with UI Type

Fail if any of the following occurs:

  • UI Type mismatch
  • Layout Paradigm mismatch
  • panorama-scene / image-led / scene-led UI is abstracted as dashboard
  • reference-visible component lacks bbox / bboxRef / visual evidence
  • Rejected Assumptions / Prohibited Components are violated
  • Asset Strategy is missing or conflicts with UI Type
  • main visual should be raster asset but is specified as CSS/SVG redraw
  • tokens.json cannot be parsed

Supporting Files

Read these references directly:

  • references/review-workflow.md
  • references/review-checklists.md
  • references/ui-type-review-rules.md
  • references/visible-evidence-rules.md
  • references/rejected-assumptions-rules.md
  • references/asset-strategy-review.md
  • references/pass-fail-rules.md
  • references/output-files.md

Examples:

  • examples/spec-review-report.example.md
  • examples/conflict-log.example.md
  • examples/revised-files-needed.example.md
  • examples/optional-patch.example.md

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Visual Spec Review AI skill do?

This skill should be used after visual-to-spec has produced 03_visual_spec/ and before design-system-governor or frontend implementation begins. It independently reviews extracted visual specs against the reference screenshot, UI type, layout paradigm, visible-element evidence, rejected assumptions, and asset strategy, catching hallucinations, inconsistencies, and implementation blockers.

Why use Visual Spec Review on TypingMind?

Because you install it once and use it with any model. Visual Spec Review is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Visual Spec Review in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Jason904/ui-skill-lab/tree/main/.codex/skills/visual-spec-review. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Visual Spec Review?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Visual Spec Review?

As many as you like. As long as a model supports skills, you can use Visual Spec Review with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Visual Spec Review AI skill free?

Yes. It is published on GitHub by Jason904 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇