diff --git a/build_tools/extract_help_sections.fish b/build_tools/extract_help_sections.fish
new file mode 100644
index 000000000..b48f837ad
--- /dev/null
+++ b/build_tools/extract_help_sections.fish
@@ -0,0 +1,23 @@
+#!/usr/bin/env fish
+# Build a list of all sections in the html sphinx docs, separately by page,
+# so it can be added to share/functions/help.fish
+# Use like
+#    fish extract_help_sections.fish user_doc/html/{fish_for_bash_users.html,faq.html,interactive.html,language.html,tutorial.html}
+# TODO: Currently `help` uses variable names we can't generate, so it needs to be touched up manually.
+# Also this could easily be broken by changes in sphinx, ideally we'd have a way to let it print the section titles.
+#
+
+for file in $argv
+    set -l varname (string replace -r '.*/(.*).html' '$1' -- $file | string escape --style=var)pages
+    # Technically we can use any id in the document as an anchor, but listing them all is probably too much.
+    # Sphinx stores section titles (in a slug-ized form) in the id,
+    # and stores explicit section links in a `span` tag like
+    # `<span id="identifiers"></span>`
+    # We extract both separately.
+    set -l sections (string replace -rf '.*class="headerlink" href="#([^"]*)".*' '$1' <$file)
+    # Sections titled "id5" and such are internal cruft and shouldn't be offered.
+    set -a sections (string replace -rf '.*span id="([^"]*)".*' '$1' <$file | string match -rv 'id\d+')
+
+    set sections (printf '%s\n' $sections | sort -u)
+    echo set -l $varname $sections
+end