Skip HTML/XML tags and non-path targets in link validation

Comrak parses inline HTML tags like <ns:tag> as autolinks. Filter
these out by skipping targets containing colons (that aren't known
URL schemes) or quotes/commas (JSON attribute garbage).

Author: nicosuave

src/links.rs

@@ -25,14 +25,38 @@ pub struct Link {
     pub len: usize,
 }
 
+const KNOWN_SCHEMES: &[&str] = &["http://", "https://", "mailto:", "tel:", "ftp://"];
+
 impl Link {
     /// Whether this is an external URL (http, https, mailto, etc.)
     pub fn is_external(&self) -> bool {
         let t = &self.raw_target;
-        t.starts_with("http://")
-            || t.starts_with("https://")
-            || t.starts_with("mailto:")
-            || t.starts_with("tel:")
+        KNOWN_SCHEMES.iter().any(|s| t.starts_with(s))
+    }
+
+    /// Whether this link should be skipped entirely during validation.
+    /// Filters out targets that are clearly not file paths or URLs,
+    /// such as HTML/XML tag names that comrak misparses as links.
+    pub fn should_skip(&self) -> bool {
+        let t = &self.raw_target;
+
+        // External URLs are handled separately
+        if self.is_external() {
+            return false;
+        }
+
+        // Contains a colon but isn't a known scheme: likely XML namespace,
+        // custom protocol, or other non-file-path content
+        if t.contains(':') {
+            return true;
+        }
+
+        // Targets with quotes/commas are JSON or attribute garbage
+        if t.contains('"') || t.contains(',') {
+            return true;
+        }
+
+        false
     }
 
     /// Decoded file target path (percent-decoded).

src/rules/broken_links.rs

@@ -19,6 +19,9 @@ impl WorkspaceRule for BrokenLinks {
 
         for file in &workspace.files {
             for link in &file.links {
+                if link.should_skip() {
+                    continue;
+                }
                 if link.is_external() && !config.links.check_external {
                     continue;
                 }

tests/integration.rs

@@ -238,6 +238,19 @@ fn test_external_links_skipped_by_default() {
         .success();
 }
 
+#[test]
+fn test_html_xml_tags_not_treated_as_links() {
+    let vault = create_vault(&[
+        ("doc.md", "# Doc\n\nSome text with <custom:tag>content</custom:tag> inline.\n\n<my-ns:element attr=\"val\">block</my-ns:element>\n"),
+    ]);
+
+    cmd()
+        .arg(vault.path())
+        .assert()
+        .success()
+        .stderr(predicates::str::contains("broken-links").not());
+}
+
 #[test]
 fn test_links_in_code_blocks_ignored() {
     let vault = create_vault(&[