EXAMPLE REPORT· fictional product · illustrative evidence · not a customer result
Example deliverable · audit v2 · 20 checks

Northstar Metrics API

A fictional B2B usage-analytics product used to show exactly what the $49 report contains: request evidence, weighted priorities, implementation guidance, the facts needed from the owner, and the work I would deliberately decline.

northstar-metrics.example audited: 2026-08-07 scope: origin + pricing page
01 / decision summary

Four useful fixes. Four checks to leave alone.

The page is readable, but its commercial facts are still implied rather than typed. The recommendation targets that gap without inventing capabilities for the sake of 100.

  • Already solid: product summary, HTTPS, robots policy, sitemap, compact llms.txt, canonical URL and an honest informational agent card.
  • Fix first: typed product/offer facts, discoverable machine twins and complete Open Graph metadata.
  • Fix second: a generated full-text llms companion from approved source pages.
  • Do not fake: MCP, Agent Skills or response-signing documents for capabilities Northstar does not operate.
D 65current
A 93expected
02 / priority queue

Evidence, impact and the exact next move.

Weight determines order, then human review separates a meaningful omission from a fashionable file the product has no reason to publish.

P1weight 15

State the product and offer as typed facts

GET /pricing → visible Pro offer at $79/month; no script[type="application/ld+json"] in the response.

An agent must infer that Northstar is software and that $79/month is an offer rather than a stray number. Generate SoftwareApplication and Offer JSON-LD from the same data that renders the visible card.

Needed from the owner: billing period, currency, availability and whether the visible price excludes tax.

P1weight 8

Advertise the machine-readable twins

The head declares a canonical and stylesheet; no rel="alternate" has an application/json or text/markdown media type.

A caller should discover structured content from the page it is evaluating, not guess the path. Add both alternates to the shared commercial template and send the equivalent Link header.

<link rel="alternate" type="application/json"
      href="https://northstar-metrics.example/product.json">
<link rel="alternate" type="text/markdown"
      href="https://northstar-metrics.example/product.md">
P1weight 7

Complete the share-card contract

og:title yes · og:description yes · og:image no

Assistant and search interfaces have no declared visual and may choose an unrelated asset. Publish one stable 1200×630 product card and add its URL, dimensions and descriptive alt text.

Needed from the owner: approve the proposed product card or provide the existing brand asset.

P2weight 5

Add the full-text llms companion

GET /llms.txt → 200 with H1 and summary · GET /llms-full.txt → 404

An agent can find the right pages but needs several more fetches to answer product, price, authentication and limits questions. Generate /llms-full.txt from approved source content; do not concatenate private, duplicate or stale documentation.

Needed from the owner: confirm which documentation pages are safe and current enough to reproduce in full.

03 / complete result

All 20 checks, including the inconvenient ones.

The score is normalized from 79 earned points across 122 possible weighted points. Nothing is hidden because it does not fit the sales narrative.

v2 checklist · fictional response evidence12 pass · 4 fix · 4 defer
checkwresultevidence / disposition
llms.txt published15pass200; H1 names Northstar Metrics API
llms.txt title + summary5passblockquote summary follows the H1
llms-full.txt published5fail · P2404; generate from approved docs
robots.txt published5pass200 text/plain
AI crawlers not blocked10passno audited AI agent is disallowed
sitemap.xml published8pass200 XML; also declared in robots.txt
schema.org JSON-LD15fail · P1no schema.org node on /pricing
title + description7passboth present and product-specific
Open Graph metadata7fail · P1image missing
canonical URL5passdeclares the audited HTTPS URL
machine alternates8fail · P1no JSON or Markdown alternate
agent card10passcurrent path; explicitly informational
HTTPS5passpage and canonical use HTTPS
Content Signals3passsearch=yes, ai-input=yes, ai-train=no
current A2A card path3pass/.well-known/agent-card.json
Markdown negotiation3passAccept: text/markdown returns Markdown
MCP server card2not applicableno MCP server; do not invent one
RFC 9727 API catalog2deferadd only with stable OpenAPI ownership
Agent Skills index2not applicableno agent task exists
Web Bot Auth directory2deferneeds signing-key storage and rotation
04 / human review

A useful report says what not to build.

The free score names every pass and failure. The paid report adds product context, implementation order and a boundary around work that would only make the number prettier.

implement now

  • JSON-LD from the visible commercial source
  • JSON and Markdown discovery links
  • one approved Open Graph product card
  • generated llms-full.txt from current public docs

decline for now

  • MCP card when no MCP endpoint exists
  • Agent Skills when no agent task exists
  • API catalog before OpenAPI ownership is stable
  • signing directory without key storage and rotation
05 / implementation order

One pass through source, template and production.

The report is ready for a developer to execute, but it marks the few facts that code cannot safely decide.

  1. Confirm commercial facts and approved source pages.
  2. Add JSON-LD and alternates at the shared template.
  3. Publish the approved 1200×630 social asset.
  4. Generate the full-text llms companion.
  5. Deploy, fetch production evidence and re-audit.
06 / owner-only input

The four things I would ask the buyer.

A real report keeps account access and business truth separate from changes that can be made safely in a repository.

Needed before implementation

These are explicit so delivery does not stall halfway through a change.

  • Confirm price currency, billing period, availability and tax wording.
  • Approve the product-card visual or provide the correct brand asset.
  • Confirm which public docs are current and safe to inline.
  • Name the repository/platform and the person who can approve deployment.
Boundary: this evaluates technical discoverability and parseability. It cannot guarantee a ranking, model mention, recommendation, citation or conversion. CDN-managed policy and dashboard-only settings remain owner actions.
One real URL is enough to start

Get this report for your site.

The $49 fixed-scope report covers one public origin and one representative commercial page, with Markdown and JSON delivery within one business day after fit and payment are confirmed.