Fix llms-txt breadcrumb leaking and link prefix

- Remove breadcrumbs from extracted HTML before Pandoc conversion,
  since Pandoc strips <nav> wrappers and loses the droppable class
- Strip ./ prefix from converted .llms.md links for cleaner output
- Fix test regex to match code annotation markers with space

Co-Authored-By: Claude Opus 4.6 <[email protected]>

Author: cscheid

src/project/types/website/website-llms.ts

@@ -131,6 +131,7 @@ function extractMainContent(doc: Document): string {
     ".sidebar",
     ".quarto-search",
     "nav.navbar",
+    ".quarto-page-breadcrumbs",
     "script",
     "style",
     "link[rel='stylesheet']",

src/resources/filters/llms/llms.lua

@@ -193,6 +193,7 @@ function Link(link)
     if link.target and (link.target:match("%.html$") or link.target:match("%.html#")) then
         link.target = link.target:gsub("%.html#", ".llms.md#")
         link.target = link.target:gsub("%.html$", ".llms.md")
+        link.target = link.target:gsub("^%./", "")
         if link.classes:includes("btn") then
             link.attr = pandoc.Attr()
         end

tests/docs/smoke-all/website/llms-txt/index.qmd

@@ -13,7 +13,7 @@ _quarto:
         - []
       ensureLlmsMdRegexMatches:
         # First array: patterns that MUST match - verify anchor links, code annotations, and conditional content
-        - ["\\[callout examples\\]\\(about\\.llms\\.md#callout-examples\\)", "#<1>", "#<2>", "Load tidyverse", "Open help for ggplot", "only for LLM consumption"]
+        - ["\\[callout examples\\]\\(about\\.llms\\.md#callout-examples\\)", "# <1>", "# <2>", "Load tidyverse", "Open help for ggplot", "only for LLM consumption"]
         # Second array: patterns that must NOT match (no .html links, no annotation UI, no hidden content)
         - ["\\.html\\)", "\\.html#", "code-annotation-anchor", "should not appear in LLM output"]
 ---