AGENT READINESS
AUDIT

The skill starts from one framing: agent readiness is accessibility plus performance plus access policy, with a Markdown and discovery layer on top. WebMCP is an optional enhancement for sites with forms or transactions, not a foundation. The audit runs its checks in a fixed order and keeps each tool's JSON for the report.

1. Lighthouse fraction. lighthouse_agentic.py reads the Agentic Browsing category through PageSpeed Insights v5 or a saved report and reproduces the Lighthouse report renderer's fraction exactly.
2. HTTP and markup checks. agentic_check.py covers server-rendered content, robots.txt groups per AI agent and Content-Signal, llms.txt, Markdown delivery, ai-catalog.json, /.well-known documents, WebMCP markup, and whether unknown URLs return a real 404.
3. Accessibility tree. agent_ux_check.py produces a local 0 to 100 heuristic, presented separately from the Lighthouse fraction.
4. WAF behaviour, only with authorization. The --ua-matrix option reports how the site treats unverified traffic carrying AI agent user agents. It is never treated as proof that the real agent is blocked.
5. Commerce handoff. On e-commerce sites a UCP profile is flagged and the depth goes to the e-commerce skill.

In a full site audit, the orchestrator now spawns the seo-agentic agent alongside seo-geo, and its findings feed the AI Search Readiness category.

Primary sources: the seo-agentic skill file, the seo-agentic agent file, and the v2.4.0 release notes.

The skill's reference was verified on 2026-09-23 against the Lighthouse 13.5.0 package and live PageSpeed Insights runs. Lighthouse labels the category as still under development and subject to change, so the numbers below are dated facts, not permanent rules.

Not-applicable, manual, and informative audits drop out of N, and every counted audit passes at a score of 0.9 or higher, so N is at most 6. The JSON category score is a weighted mean, not the fraction. The skill therefore reports X/N with the Lighthouse version and form factor, never as a percentage, and lists the changes that would add a counted audit as options rather than goals. A higher N is not a better site.

Commands

Step 1: Install or update Claude SEO

/plugin marketplace add AgriciDaniel/claude-seo
/plugin install claude-seo@agricidaniel-claude-seo
/seo setup
/seo doctor

Manual installs are covered in the installation guide.

Step 2: Run the audit

/seo agentic https://your-site.com

If PageSpeed Insights is out of quota or no key is configured, the skill says so and continues with the HTTP, markup, and accessibility checks. It can also read a local Lighthouse report that you produce yourself.

Step 3: Review the drafts before changing anything

Fix mode prints drafts to the terminal. The robots draft adds Content-Signal to each group and never changes Allow or Disallow lines. The WebMCP draft proposes one tool per real form that submits through the form's own handler. Sends, purchases, and deletes stay marked consequential with a human confirmation step.

Honest-reporting rules

• No promise of ranking, citation, or traffic gains from any item.
• No vendor token-savings figures presented as independent evidence.
• No claim that a named consumer agent requests Markdown or reads llms.txt.
• Google-Extended, Content-Signal, and llms.txt are never presented as affecting Google Search.
• Training, search, and user-triggered agent access are reported on separate lines.
• WebMCP, Content-Signal, ai-catalog.json, and Web Bot Auth are labelled as drafts or proposals with the date checked.