[agent-ptolemy]

Automated fixes from the monthly SEO audit (13 October 2026).

Fixes applied

• Missing robots directive on 59 pages. overrides/main.html only emits a <meta name="robots"> tag when meta.robots is explicitly set in frontmatter — it has no implicit default. 59 of 61 canonical content pages had no robots field at all, so they rendered with no robots meta tag whatsoever. Added robots: "index, follow" to all of them, matching the convention already used correctly on 404.md (noindex, nofollow). LICENSE.md and NOTICE.md are untouched — they carry no frontmatter by design.

Validated with the full deploy pipeline after the fix: unit tests, validate_headings.py, zensical build --strict, breadcrumb enrichment, sitemap + updates.json generation, validate_seo.py, and validate_agent_readiness.py all pass clean.

Full audit results (8 checks)

1. Technical SEO — robots.txt at repo root is correct: allows all crawlers, lists all 12 required AI crawlers, references the sitemap. Sitemap has all 61 canonical pages; the 21 URLs without <lastmod> are exactly the legal pages (LICENSE/NOTICE, no frontmatter) plus the 19 pages marked review-status: "pending" — this is the template's intended behaviour (freshness signals withheld until substantive human review), not a bug.
2. Broken links — Internal: 0 broken links found across all markdown files in docs/. External: could not be checked this cycle — the execution environment's egress policy blocks outbound HTTPS to essentially all external domains (confirmed against several .gov.au hosts and even wikipedia.org, both via curl and WebFetch; only a short allowlist like github.com resolves). This is an environment limitation, not a site finding — same recurring issue as the last several audits. 534 unique external URLs are referenced across the site (197 of them .gov.au/.gov), unchecked this cycle.
3. Internal linking — the site's real navigation comes from zensical.toml (per CLAUDE.md), not in-body cross-links, so most pages show "zero incoming body links" by design; this isn't a discoverability problem (the sidebar nav links every page from every page). In-body contextual cross-linking is sparse, which is a content opportunity, not a defect — flagging for awareness only, no action taken.
4. On-page SEO — validate_seo.py passes with 0 errors (title ≤60 chars, description ≤165 chars, uniqueness, etc. all satisfied). author, og_title, og_description, og_url, og_image, twitter_card are correctly left to the template's defaults per CLAUDE.md ("The template supplies the site author, canonical URL, default social image and Twitter card type"). Many page titles/descriptions sit outside the audit checklist's 50-60/150-160 "ideal" character range, but this is a consistent, deliberate site-wide style, and rewriting them needs keyword research — reporting only, not auto-fixed.
5. Schema markup — homepage (WebSite+Organization), about page (Organization) and glossary (DefinedTermSet) all correct. All business-resources and safety-standards pages correctly render Article+BreadcrumbList via the template, and the two tagged with FAQ frontmatter render FAQPage. Gap (not auto-fixed, needs content expertise): governance-templates/ai-readiness-checklist.md and ai-vendor-evaluation-checklist.md are missing howto frontmatter (both already have faq). policy-template-library.md lacks both howto/faq too, but on inspection it's a landing/index page linking out to the other templates, not a step-by-step procedure — HowTo/FAQ schema isn't actually appropriate there, so it isn't counted as a gap.
6. Content freshness — 0 pages are overdue against their review-cycle threshold. 19 pages are flagged review-status: "pending" (substantive human review pending) — same list as before, no change this cycle.
7. Agent-readable resources — validate_agent_readiness.py passes clean. llms.txt/llms-full.txt content inventory still matches the current 11 governance templates + library index. llms.txt's Last updated: 2026-08-23 (~51 days) hasn't moved recently but no structural drift was found against it.

Issues requiring human action

• Recurring, now 3rd consecutive cycle unaddressed: the custom 404 page's nav/asset/favicon links (href="../assets/...", href="../favicon.ico", etc.) are still relative paths built for one directory level deep, while deploy.yml's "Prepare GitHub Pages 404" step serves the file from site-root /404.html without rewriting them — so they resolve one level too high for most real 404 hits. First flagged 11 Oct, still present in today's build. This is a deploy-pipeline/template change, not page content, so it's out of this PR's auto-fix scope — recommend either rewriting site/404.html's links to root-relative paths as a post-processing step in deploy.yml, or adding <base href="/"> to the 404 page's <head> in overrides/main.html.
• Write howto frontmatter for ai-readiness-checklist.md and ai-vendor-evaluation-checklist.md.
• NOTICE.md remains an orphan page (0 inbound links from nav, footer or body content) — unchanged across several consecutive audits.
• External link verification couldn't run this cycle (environment egress policy) — worth a manual pass, especially the 197 .gov.au links, given how often those restructure.
• Optional: tighten title/description lengths to the 50-60 / 150-160 char "ideal" range where it doesn't hurt clarity (not urgent — nothing here fails validation).

Source: agent-engine/maintenance/seo-report-2026-10-13.md