What is the fixing-metadata skill?
The fixing-metadata skill is a SKILL.md from the ibelick/ui-skills repo. It makes your coding agent audit page titles, meta descriptions, canonical URLs, Open Graph and Twitter cards, favicons, JSON-LD and robots directives, fix duplicates and indexing problems first, and keep every diff scoped to metadata only.
fixing-metadata is an agent skill from the ibelick/ui-skills repo. Its SKILL.md tells a coding agent to audit the metadata in a page's head, from titles and canonical tags to social cards and structured data, and to fix problems with minimal changes.
Rules are ranked in eight groups. Correctness and duplication come first, then title and description, canonical and indexing, and social cards. Icons, structured data and locale follow, and tool boundaries are marked critical.
The five-step workflow
- Find pages with missing or incorrect titles, descriptions, canonical and Open Graph tags
- Audit against the priority rules and fix critical issues, duplicates and indexing, first
- Make title, description, canonical and
og:urlagree with each other - Check social cards on a real URL, not localhost
- Keep diffs minimal and scoped to metadata only
What changes in your agent's output
- One metadata definition per page, with no duplicate title, description, canonical or robots tags
- Absolute URLs for Open Graph and Twitter images, and an
og:urlthat matches the canonical noindexon private, duplicate or non-public pages, and by default on previews and staging- JSON-LD only where it maps to content the page really renders
- A fix written in the project's own pattern, such as the Next.js metadata API, react-helmet or a manual head
Who it is for
Best for
- Shipping new marketing pages or shareable links
- Fixing a social preview that shows the wrong title or image
- Setting up canonical URLs, robots rules and favicons
- Building a shared SEO component or layout metadata defaults
Not for
- Keyword research or content strategy. The skill covers tags, not what to write in them
- Framework or SEO library migrations. It tells the agent to follow your existing pattern
- Sitemaps. They are not part of its rule set
Use it with UiChemy
After UiChemy AI to WordPress brings a build into Elementor or Gutenberg, the document head usually comes from your theme or an SEO plugin. Run the skill on that metadata, then confirm the rendered page outputs one title and one canonical. Pair it with the fixing-accessibility skill and the Impeccable polish skill for a full pre-launch pass.
How to install Fixing Metadata
git clone https://github.com/ibelick/ui-skills.gitClone the repo
Run the command above. The skill lives in
skills/fixing-metadataand is oneSKILL.mdfile.Copy the folder
Copy
skills/fixing-metadatainto your agent's skills directory. For Claude Code that is~/.claude/skills/for every project or.claude/skills/inside one project.Ask for the audit
Prompt the agent, for example: Use fixing-metadata to audit the metadata in my root layout and fix duplicates first.
Check that it loaded
Start a new session and ask the agent to use fixing-metadata on a layout file. It should list issues in priority order and propose metadata-only diffs.
Key features
- Duplicate and stability rulesMetadata lives in one place per page, with no competing systems and no duplicate title, description, canonical or robots tags. Values must be deterministic, dynamic strings are escaped, and every page keeps safe title and description defaults.
- Title and descriptionEvery page needs a title in one consistent format. Titles stay short and readable without stuffing, shareable pages get a plain-text meta description and descriptions carry no markdown or quote spam.
- Canonical and indexingThe canonical points to the preferred URL.
noindexis for private, duplicate or non-public pages only, robots meta must match access intent, previews and staging default tonoindexand paginated pages need correct canonical behavior. - Open Graph and Twitter cardsShareable pages set the Open Graph title, description and image, and both image tags use absolute URLs.
og:urlmatches the canonical,og:typeis usually website or article andtwitter:carddefaults tosummary_large_image. - Icons and manifestAt least one favicon must work across browsers, with an apple-touch-icon where relevant. A manifest must be valid and referenced,
theme-coloris set on purpose, and icon paths stay stable and cacheable. - Structured data disciplineJSON-LD is added only when it maps to real page content and must be valid. The agent never invents ratings, reviews, prices or organization details, and prefers one block per page.
- Locale and alternatesThe
htmllang attribute is correct,og:localeis set when localization exists and hreflang alternates are added only for pages that really exist. Localized pages canonicalize per locale. - Four-way agreement checkTitle, description, canonical and
og:urlmust all agree. The workflow lists this as its own step, and the review guidance repeats it. - Real-URL verificationSocial cards are verified on a real URL, not localhost. The workflow and the review guidance both repeat this, so the agent checks the card where a platform would actually fetch it.
- Tool boundariesChanges stay minimal and scoped to metadata. The agent does not migrate frameworks or SEO libraries, and it follows the project's existing metadata pattern.
Frequently asked questions
What is the fixing-metadata skill?
It is a SKILL.md from the ibelick/ui-skills repo that teaches a coding agent to audit and fix HTML metadata. It covers page titles, meta descriptions, canonical URLs, Open Graph tags, Twitter cards, favicons, JSON-LD structured data and robots directives, and it keeps every diff scoped to metadata only.
How do I install the fixing-metadata skill?
Clone the ibelick/ui-skills repo with git, then copy the skills/fixing-metadata folder into your agent's skills directory. In Claude Code that is ~/.claude/skills/ for all projects or .claude/skills/ inside one project. The folder contains a single SKILL.md, so nothing else needs installing.
What does the fixing-metadata skill check first?
Correctness and duplication come first. The agent looks for competing metadata systems, duplicate title, description, canonical or robots tags, unstable values and missing defaults. Title and description, canonical and indexing, and social cards rank high, and the workflow says to fix duplicates and indexing problems before anything else.
Does fixing-metadata fix Open Graph and Twitter card problems?
Yes. Shareable pages must set an Open Graph title, description and image. Image URLs must be absolute, og:url must match the canonical and twitter:card defaults to summary_large_image. The agent should verify the card on a real URL, because localhost previews do not show what social platforms will fetch.
Will fixing-metadata invent structured data for my pages?
No. The rules say not to add JSON-LD unless it clearly maps to real page content, to keep it valid and to reflect what is actually rendered. The agent must never invent ratings, reviews, prices or organization details, and it prefers one structured data block per page.
Does fixing-metadata work with Next.js or react-helmet?
The skill says to follow the project's existing metadata pattern and names the Next.js metadata API, react-helmet and a manual head as examples. It does not tell the agent to switch frameworks or SEO libraries, so fixes land in the system your project already uses.
Is the fixing-metadata skill free to use?
Yes. The skill's own frontmatter lists the MIT license, and the whole ibelick/ui-skills repo is published under MIT. The skill is a plain text SKILL.md file, and installing it means copying one folder from the repository into your agent's skills directory.
Which other skills are in the same repo?
The ibelick/ui-skills repo also holds baseline-ui, create-design-md, fixing-accessibility, fixing-motion-performance and improve-ui. Pair fixing-metadata with fixing-accessibility for ARIA and keyboard issues, or with fixing-motion-performance for animation jank. Each skill is a separate folder, so you can install only the ones you need.