How to use the Broken External Links Checker agent skill

What the skill makes the agent do

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. This one covers outbound links: the agent is asked to scan a page for external links that lead to 404 pages or other error states, update each broken link to a working URL or remove it if the resource no longer exists, and avoid linking to expired domains that may have been repurposed.

The skill also explains why this matters. Broken links create a poor user experience and can signal to search engines that content is outdated or poorly maintained. For code reviews, the skill asks the agent to check rendered HTML, structured data and response headers, and to verify the final page output rather than trusting source files alone.

• Finds external links returning 4xx or 5xx status codes
• Suggests replacing each broken link with a current URL or removing it
• Checks rendered HTML and HTTP responses, not only source files

Install the skill from the ZIP download

Download the ZIP for this skill from this page and unzip it. You can then place the folder in one of two locations. For a personal setup that works across projects, put it in ~/.claude/skills/. To limit it to a single project, put it in that project's .claude/skills/ folder.

If you use the Claude apps instead of a local setup, you can upload the skill under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/broken-external-links/
• Project only: .claude/skills/broken-external-links/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

Trigger the skill with natural requests

You do not need a special command. The agent decides to use the skill when your request matches its description, which covers auditing outbound links and fixing or removing broken external links. Name the page or files and say what you want done. Clear requests work best, such as asking for a review before launch or a cleanup of an older article.

The skill's description also mentions metadata, crawlability, structured data and indexability. Requests on those topics may bring it into play, so mention link checking explicitly if that is your goal.

• "Check the external links in content/blog/launch-post.md and fix any that return 404."
• "Review this page for broken outbound links before we publish. Remove any that are dead."
• "Audit the references in our docs folder for link rot and suggest replacements."

What the skill needs and how it works

The skill lists no required tools or packages. It does not run scripts. It does use the network, because checking whether a link is broken means requesting the target URL. The skill's text is mostly guidance, and the detailed implementation is in references/rule.md, which includes an HTML example of good and bad link usage. The skill itself does not include a crawler, so the agent must build or run its own checking method.

Expect the agent to rely on its own tools to make HTTP requests. Ask it to report which URLs it checked and what status each returned, so you can confirm the results yourself.

• Files: SKILL.md and references/rule.md
• Runs scripts: no
• Uses network: yes
• Estimated time in the metadata: 10 minutes, at intermediate difficulty

Safety notes on scripts and network use

The skill does not run scripts, but the agent will send HTTP requests to third-party servers when it checks links. Only ask for checks on sites you are allowed to audit, and be aware that many requests to other sites can be slow and may be rate-limited or blocked.

Because the skill has no crawler, scanning large sites means the agent must write its own tooling. Review any code it produces before running it. Links loaded by JavaScript or hidden behind logins may be missed, so do not treat a clean result as a guaranteed complete audit.

• Only audit sites and servers you have permission to query
• Expect slowdowns or blocks on large scans
• Review agent-written crawling code before running it
• Treat results as likely but not guaranteed complete

Tips for better results

Give the agent a bounded scope, such as one article or one folder, rather than an entire site. Ask it to verify the rendered page and the HTTP response, since the skill recommends that over reading source files. When a link is dead, ask for a replacement from a live source or for the link to be removed, and check that replacements actually resolve.

• Start with one page to see how the agent reports results
• Ask for the status code of each link checked
• Check replacement URLs yourself before publishing