How the BridgeToAgent kit maps to every Lighthouse Agentic Browsing audit
Lighthouse Agentic Browsing has nine audits. The kit closes six of them outright (llms-txt-present, llms-txt-well-formed, agents-json-present, agents-json-actions-typed, agent-runbook-present, auto-discovery-links). The other three are CMS- or theme-side and the kit reports them rather than faking a fix. Here's the audit-by-audit map.
How the BridgeToAgent kit maps to every Lighthouse Agentic Browsing audit
The Chrome team's Lighthouse Agentic Browsing category scores nine specific audits. They group naturally into file-presence, file-correctness, discovery, content-structure, and one emerging spec. If you've run the audit and seen failures, the per-audit table below is the honest map of what a $49 kit closes for you versus what the kit can only report (because it depends on your CMS, your theme, or a spec that hasn't settled yet).
The point of this post isn't to claim the kit fixes everything. It's to be precise about what it fixes so the buyer evaluating whether to ship can decide if the six audits the kit closes are worth the $49 — they almost always are — and what additional work remains in their hands afterward.
The nine audits at a glance
| # | Audit | What it checks | Kit closes? |
|---|---|---|---|
| 1 | llms-txt-present | Is /llms.txt reachable with a plain-text content type? | ✓ Yes |
| 2 | llms-txt-well-formed | Does /llms.txt parse against the llmstxt.org reference parser? | ✓ Yes |
| 3 | agents-json-present | Is /agents.json reachable as valid JSON? | ✓ Yes |
| 4 | agents-json-actions-typed | Does every action in /agents.json have typed parameters? | ✓ Yes |
| 5 | agent-runbook-present | Is /agent-instructions.md reachable? | ✓ Yes |
| 6 | auto-discovery-links | Does the homepage <head> contain three <link rel="alternate"> tags pointing at the kit files? | ✓ Yes |
| 7 | sitemap-discoverable | Is /sitemap.xml valid and referenced from /robots.txt? | ✗ CMS-side |
| 8 | schema-org-density | Are there Schema.org JSON-LD blocks at minimum density on the homepage and primary pages? | ✗ Theme-side |
| 9 | webmcp-annotations | Are there WebMCP per-element annotations on interactive elements? | ✗ Spec still moving |
Six of nine in one $49 install. The remaining three are walked through below — what they actually score, why the kit doesn't try to ship them, and where the fix lives instead.
Audit 1 — llms-txt-present
What it checks. Whether /llms.txt is reachable at your domain root with a text/plain content type. Returns pass/fail.
What the kit ships. A generated llms.txt file populated from a real sitemap-driven crawl of your site. Sections are populated with the canonical URLs that matter on your domain, each annotated with a one-line summary so the model picks the right URL on the first read. Generated in two minutes from a real crawl, not a template.
Quick-fix path without the kit. Hand-write the file per the llmstxt.org spec, upload to your domain root, ensure the content type is text/plain.
Audit 2 — llms-txt-well-formed
What it checks. Whether /llms.txt parses against the llmstxt.org reference parser. This is a stricter check than presence — many hand-written and tool-generated files pass the presence check and fail the format check (malformed headers, missing section structure, broken Markdown links).
What the kit ships. Validation against the reference parser is a build-time gate. A generated file that doesn't parse cleanly doesn't ship — the audit pipeline blocks it before delivery. Files that ship pass this audit by construction.
Quick-fix path without the kit. Validate your hand-written file against the llmstxt.org reference parser — paste in the file content, fix anything the parser flags. Re-validate after every meaningful edit.
Audit 3 — agents-json-present
What it checks. Whether /agents.json is reachable as valid JSON at your domain root. Returns pass/fail.
What the kit ships. A generated agents.json with the canonical actions on your site (browse_products, search, contact, view_policy, book_demo, etc.) populated from form-extraction over the rendered DOM of your sitemap. Every action's typed parameters are validated against the action manifest reference. Every URL in the file is link-resolved against your live site before delivery.
Quick-fix path without the kit. Hand-write the JSON per the agents.json reference, ensure all action URLs resolve, ensure every parameter is typed (no null parameter blocks).
Audit 4 — agents-json-actions-typed
What it checks. Whether every action declared in /agents.json has a typed parameter schema — no null parameter blocks, no missing field types, no untyped optional fields. The hardest of the file-correctness audits because most hand-written agents.json files fail it.
What the kit ships. Typing is mandatory in the generator pipeline. The kit's action-extraction passes find forms and links, infer parameter types from input attributes (<input type="email"> → string with email format; <input type="number"> → number; <select> → enum, etc.), and refuse to emit untyped parameters. Files that ship pass this audit by construction.
Quick-fix path without the kit. Audit your hand-written agents.json for any parameters: null or missing-type-field entries. Fill them in with the actual types your endpoints expect.
Audit 5 — agent-runbook-present
What it checks. Whether /agent-instructions.md is reachable at your domain root.
What the kit ships. A generated agent-instructions.md with a navigation guide (which URLs to read for which kinds of questions), behavioral rules (don't quote prices excluding VAT, refund policy lives at /policies/refund and is canonical, etc.), and an escalation path (how to reach a human if the agent is stuck). The runbook is what tells agents how to behave on your site, not just how to read it.
Quick-fix path without the kit. Write a markdown file covering the agent navigation guide, your behavioral rules for AI handlers, and contact paths for escalation. Upload to your domain root.
Audit 6 — auto-discovery-links
What it checks. Whether the homepage <head> contains three <link rel="alternate"> tags pointing at the kit files. Specifically:
<link rel="alternate" type="application/json" title="Agent Action Map" href="/agents.json">
<link rel="alternate" type="text/plain" title="LLM Context" href="/llms.txt">
<link rel="alternate" type="text/markdown" title="Agent Runbook" href="/agent-instructions.md">
Why it matters: agents that don't probe /llms.txt paths blindly look at the <head> for discovery hints. The auto-discovery tags are the spec-aligned way to surface the kit files to those agents.
What the kit ships. The exact three-line HTML snippet is included in the kit ZIP with platform-specific install instructions for where to paste it (Shopify theme.liquid, WordPress header.php or "Insert Headers and Footers", Webflow Custom Code → Head, Next.js app/layout.tsx, etc.). Six lines of Liquid / HTML / TSX depending on your platform — your CMS handles the rest.
Quick-fix path without the kit. Copy the three <link rel="alternate"> tags into your homepage <head>. Single edit per site.
Audit 7 — sitemap-discoverable (CMS-side)
What it checks. Whether /sitemap.xml is valid and referenced from /robots.txt via a Sitemap: directive. This is a baseline SEO + discovery hygiene check — agents enumerate your URL surface via the sitemap when they're not crawling JavaScript.
Why the kit doesn't try to fix this. Your CMS owns sitemap.xml and robots.txt generation. Shopify, WordPress (via Yoast / RankMath), Webflow, Squarespace, Wix, Ghost, Next.js (via next-sitemap) all emit a sitemap by default and either auto-add or document the robots.txt reference. Generating an external sitemap from the kit would conflict with what your CMS ships.
What the kit does instead. The free audit at the kit homepage reports whether your robots.txt includes the Sitemap: line and flags it if missing. The remediation is a one-line edit in your CMS's robots.txt configuration — Yoast/RankMath dashboards, Shopify's automated robots.txt, or your next-sitemap.config.js.
Audit 8 — schema-org-density (theme-side)
What it checks. Whether your homepage and primary product / article / FAQ pages emit Schema.org JSON-LD blocks at minimum density. Specifically: an Organization block sitewide, plus page-type Schema (Article for blog posts, Product for product pages, FAQPage for FAQ pages).
Why the kit doesn't try to fix this. Schema is theme-emitted, not file-uploaded. Adding a JSON-LD block to your homepage requires either editing your theme's template files or installing a Schema plugin in your CMS — neither is a $49-deliverable. The depth of Schema you need also depends on your business model (e-commerce stores need Product Schema; SaaS sites need SoftwareApplication Schema; service businesses need LocalBusiness Schema), and the kit doesn't know which one you are.
What the kit does instead. The audit reports schema-org-density failures per-page and points you at the platform-specific fix path: modern Dawn-derived Shopify themes ship Schema natively; Yoast / RankMath / SEOPress on WordPress; theme-native on Webflow; structured-data fields on Squarespace 7.1. Per-platform install pillars (Shopify, WordPress, Webflow, Squarespace+Wix) walk this surface in detail.
Audit 9 — webmcp-annotations (spec still moving)
What it checks. Whether interactive elements (<button>, <form>, <input>) have WebMCP per-element annotations declaring them as agent-callable actions. The WebMCP spec went live in Chrome 146 (Feb 2026); Edge 147 followed in March.
Why the kit doesn't ship this in 2026. The annotation patterns aren't settled. No major framework, CMS, or theme ships WebMCP-annotated output natively. The audit is intentionally low-weighted inside the Lighthouse category precisely because the spec is still moving. Hand-annotating a theme today is work that doesn't compound — the annotation conventions will shift before they stabilize.
What's coming. The kit plans WebMCP-annotated output as a 1.3-version add-on when the annotation conventions stabilize. Existing customers get the upgraded files via re-download.
The honest summary
| Surface | Kit closes | What's left for you |
|---|---|---|
| File presence (3 audits) | All three | Nothing |
| File correctness (2 audits) | Both | Nothing |
| Discovery (2 audits) | auto-discovery-links ✓ | sitemap-discoverable is one robots.txt line in your CMS |
| Content structure (1 audit) | None directly | schema-org-density is your theme / Schema plugin — fix path documented per platform |
| Emerging (1 audit) | None | webmcp-annotations — wait for the spec to settle, then the kit 1.3 upgrade |
Six audits closed by a $49 install. One audit (the Sitemap: line) is a one-line config edit. One audit (Schema) is platform-specific and documented in the platform pillar for your CMS. One audit (WebMCP) nobody passes in 2026 and nobody is expected to.
The kit's scope is the spec-aligned file layer. The complementary surfaces (sitemap discovery, Schema density) sit in your CMS where they're better solved natively. The intentional scope-bounding is what keeps the kit at $49 and the install at two minutes — anything broader would be an integration project, not a kit.
Related
- Lighthouse Agentic Browsing — every audit, every fix → — the canonical per-audit fix reference (companion to this post; this one is what-the-kit-closes, that one is the universal fix recipe)
- We measured Lighthouse Agentic Browsing scores on 11 sites before and after kit install → — real before/after cohort data backing the "6 of 9" claim with measured per-audit closure rates
- Cloudflare's Agent Readiness Score vs Chrome Lighthouse → — the comparison pillar that frames why Lighthouse is the audit to optimize for first
- What Cloudflare's four dimensions actually check → — the working-depth deep-dive on the alternative scoring framework
- Agent-readiness scoring frameworks reference → — Lighthouse, Cloudflare, Bridge AI, The Optimisers side-by-side
- The 2026 agent-ready Shopify checklist → — platform-specific application
- The 2026 agent-ready WordPress checklist → — platform-specific application
- The 2026 agent-ready Webflow checklist → — platform-specific application
- The 2026 agent-ready Squarespace and Wix checklist → — platform-specific application
- Agentic Kit install matrix — all 13 platforms side-by-side → — once you've decided the audits are worth closing, pick the install path for your platform
- Run the free readiness audit → — five-second check that replicates the public Lighthouse signals