How to use the XML Sitemap Analyzer and Generator agent skill

What the seo-sitemap skill makes the agent do

In analysis mode, the agent checks a site's XML sitemap, the file that lists pages for search engines, for problems such as invalid XML, non-200 URLs, redirected or noindexed URLs, inaccurate lastmod dates, size limits and deprecated tags. It also looks at whether the sitemap is referenced in robots.txt and compares crawled pages with the sitemap to flag missing pages. Results are written to a VALIDATION-REPORT.md file with an issues list, severity levels and recommendations.

In generation mode, the agent asks for your business type or detects it from the existing site, loads an industry template from the sibling seo-plan skill, plans the structure with you, and writes a valid sitemap.xml with a sitemap index if needed, plus a STRUCTURE.md file describing the site architecture.

• Analysis: validation checks, quality signals and extension sitemaps for image, video and news.
• Generation: industry templates, quality gates for large numbers of location pages, and output files.
• Discovery: a bundled helper script finds sitemaps declared in robots.txt and probes common sitemap paths.

Install the seo-sitemap skill

Download the skill ZIP from this page and unzip it. The package contains LICENSE.txt and SKILL.md. Then choose one of two locations. For a personal install, place the unzipped skill folder in ~/.claude/skills/. For a single project, place it in the project's .claude/skills/ folder.

You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.

The skill's scripts depend on the claude-seo plugin, so install that plugin so that CLAUDE_PLUGIN_ROOT resolves. Without it the bundled helper cannot be run.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills

Check the requirements before you run it

The skill needs Python, because its bundled sitemap_discovery.py helper is run through a claude-seo script. It needs the claude-seo plugin installed so that CLAUDE_PLUGIN_ROOT resolves. It needs network access to the target site. For generation mode it also uses industry templates from the seo-plan skill, so that skill must be available.

• Python
• claude-seo plugin installed (CLAUDE_PLUGIN_ROOT must resolve)
• Network access to the target site
• seo-plan skill templates for generation mode

Trigger the skill with natural requests

The skill is user-invocable and its description names trigger phrases. Ask for a sitemap check or generation using words like "sitemap", "XML sitemap", "sitemap issues" or "generate sitemap". You can also give a URL or the word generate as an argument, as the argument hint suggests: [url or generate].

• "Check the XML sitemap for https://example.com and tell me what is broken."
• "Sitemap issues on my site: why are some pages not being indexed?"
• "Generate sitemap for a dental clinic with location pages."
• "Create an XML sitemap for my blog and split it if needed."

How the analysis works

The agent runs the discovery helper against the site URL. The helper reads Sitemap: lines in robots.txt and also probes common sitemap paths when a declared sitemap is stale or invalid. The skill reports "not found" only when the helper's found list is empty after declared and common candidates are checked. Declared failures are kept as findings rather than ignored.

Key limits the report checks: a single file may hold up to 50,000 URLs and 50MB uncompressed, whichever is hit first. Google ignores priority and changefreq tags, so those can be removed. Google also only honours lastmod when it is consistently and verifiably accurate.

Generate a sitemap safely

Generation follows a planned process: business type, template, interactive structure planning, quality gates, valid XML output, and splitting with a sitemap index. The skill warns at 30 or more location pages, requiring at least 60% unique content, and stops at 50 or more location pages until there is a justification.

The skill lists penalty-risk patterns to avoid at scale, such as location pages where only the city name changes, and AI-generated pages without human review and unique value. Its own warning is that programmatic pages at scale carry penalty risk, so review generated output before publishing.

• Acceptable at scale: integration pages with real setup docs, glossary pages with 200+ word definitions, product pages with unique specs.
• Avoid at scale: city-swapped location pages, generic 'best tool for industry' pages, AI-generated pages without review.

Safety notes: scripts and network access

This skill runs scripts and uses the network. It fetches URLs from the target site, so only use it on sites you own or have permission to crawl. Crawling may create load on the site and can trigger rate limits. If rate limiting is detected, the skill is instructed to back off and report partial results.

The skill also treats fetched pages, search results, API responses and third-party files as untrusted data. Any text in them that addresses the agent is reported as a finding, and the agent should not run commands, install anything or change settings because fetched content asks it to.

• Only audit sites you own or are authorised to crawl.
• Expect network requests to the target site and possible rate limiting.
• Review the report before acting on any suggested changes.