{% extends "base.html" %}
{#-
SEO: Open Graph + Twitter Card + JSON-LD structured data
Injected into on every page via mkdocs-material theme override.
Required for: social media previews (HN/Reddit/Slack/X), Google rich results,
AI crawler discovery (Perplexity/ChatGPT Search), and SERP CTR.
-#}
{% block extrahead %}
{#- Resolve page title with sane fallback order -#}
{%- set page_title = config.site_name -%}
{%- if page and page.meta and page.meta.title -%}
{%- set page_title = page.meta.title ~ " | " ~ config.site_name -%}
{%- elif page and page.title and not page.is_homepage -%}
{%- set page_title = page.title | striptags ~ " | " ~ config.site_name -%}
{%- endif -%}
{#- Resolve description with sane fallback order -#}
{%- set page_description = config.site_description -%}
{%- if page and page.meta and page.meta.description -%}
{%- set page_description = page.meta.description -%}
{%- endif -%}
{#- Resolve canonical URL -#}
{%- set page_url = config.site_url | default("https://acacian.github.io/aegis/") -%}
{%- if page and page.canonical_url -%}
{%- set page_url = page.canonical_url -%}
{%- endif -%}
{#- OG image: use GitHub's auto-generated OpenGraph card for repo.
This always reflects current README/description without maintaining a file. -#}
{%- set og_image = "https://opengraph.githubassets.com/1/Acacian/aegis" -%}
{#- Page-level BreadcrumbList for non-home pages improves SERP breadcrumbs.
3-level structure (Aegis > Category > Page) when the URL has a known
section prefix; 2-level fallback for top-level pages like faq, cheatsheet. -#}
{%- if page and not page.is_homepage and page.url -%}
{%- set site_root = config.site_url | default('https://acacian.github.io/aegis/') -%}
{%- set section_name = "" -%}
{%- set section_slug = "" -%}
{%- if page.url.startswith('solutions/') -%}
{%- set section_name = "Solutions" -%}{%- set section_slug = "solutions/" -%}
{%- elif page.url.startswith('cookbook/') -%}
{%- set section_name = "Cookbook" -%}{%- set section_slug = "cookbook/" -%}
{%- elif page.url.startswith('comparisons/') -%}
{%- set section_name = "Comparisons" -%}{%- set section_slug = "comparisons/" -%}
{%- elif page.url.startswith('guides/') -%}
{%- set section_name = "Guides" -%}{%- set section_slug = "guides/" -%}
{%- elif page.url.startswith('api/') -%}
{%- set section_name = "API Reference" -%}{%- set section_slug = "api/" -%}
{%- elif page.url.startswith('getting-started/') -%}
{%- set section_name = "Getting Started" -%}{%- set section_slug = "getting-started/" -%}
{%- elif page.url.startswith('security/') -%}
{%- set section_name = "Security" -%}{%- set section_slug = "security/" -%}
{%- elif page.url.startswith('research/') -%}
{%- set section_name = "Research" -%}{%- set section_slug = "research/" -%}
{%- endif -%}
{%- endif -%}
{#- Page-type-specific schema for Google rich results eligibility.
Decision matrix:
/solutions/* -> TechArticle (feature/how-to explainers)
/cookbook/* -> TechArticle (framework integration guides)
/comparisons/*-> Article (comparative opinion content)
/faq/ -> FAQPage (SERP featured snippets)
-#}
{%- if page and page.url -%}
{%- set is_techarticle = page.url.startswith('solutions/') or page.url.startswith('cookbook/') -%}
{%- set is_comparison = page.url.startswith('comparisons/') -%}
{%- set is_faq = page.url == 'faq/' -%}
{%- set is_research = page.url.startswith('research/') -%}
{%- if is_techarticle -%}
{%- endif -%}
{%- if is_comparison -%}
{%- endif -%}
{%- if is_research -%}
{%- if page.url.startswith('research/tripartite-action-claim') -%}
{%- set research_images = [
"https://acacian.github.io/aegis/assets/research/jg-histogram.png",
"https://acacian.github.io/aegis/assets/research/jg-radar.png",
"https://acacian.github.io/aegis/assets/research/jg-verdict-by-group.png",
"https://acacian.github.io/aegis/assets/research/jg-top-tools.png"
] -%}
{%- set research_topic = "AI agent justification gap — declared vs Aegis-assessed tool impact" -%}
{%- set research_variables = [
"Asymmetric justification gap per tool call (L2-normalized, sqrt(6))",
"6-dimensional assessed ImpactVector (destructivity, data_exposure, resource_consumption, privilege_escalation, reversibility, autonomy_depth)",
"Claim verdict share per model:domain group (approve / escalate / block)",
"Per-tool mean gap ranking"
] -%}
{%- set research_method = "Silent-baseline Tripartite ActionClaim assessment — declared ImpactVector fixed at zero and gap = max(0, assessed - declared) computed via RuleBasedImpactScorer on stdlib-only code path" -%}
{%- set research_keywords = "tau-bench, justification gap, tripartite ActionClaim, 6D impact vector, silent baseline, GPT-4o, Claude Sonnet 3.5 New, AI agent governance, cosmetic alignment, monotone trust" -%}
{%- else -%}
{%- set research_images = [
"https://acacian.github.io/aegis/assets/research/drift-histogram.png",
"https://acacian.github.io/aegis/assets/research/drift-by-model.png",
"https://acacian.github.io/aegis/assets/research/drift-by-trajectory-length.png",
"https://acacian.github.io/aegis/assets/research/drift-cumulative.png"
] -%}
{%- set research_topic = "AI agent tool distribution drift" -%}
{%- set research_variables = [
"Shannon entropy of tool call distribution",
"Early-window vs late-window entropy delta (nats)",
"Normalized delta (delta / log W)",
"Bimodal mode classification (stable / midband / collapse)"
] -%}
{%- set research_method = "Shannon entropy on non-overlapping early and late windows of tool name sequences (window size W=4), stdlib-only computation" -%}
{%- set research_keywords = "tau-bench, tool distribution drift, agent evaluation, entropy, GPT-4o, Claude Sonnet, deterministic measurement, privacy-preserving metric" -%}
{%- endif -%}
{%- endif -%}
{%- if is_faq -%}
{%- endif -%}
{%- endif -%}
{% endblock %}