diff --git a/src/parser/org.rs b/src/parser/org.rs
index 9019f58..d5ae880 100644
--- a/src/parser/org.rs
+++ b/src/parser/org.rs
@@ -167,7 +167,8 @@ struct OpenGreater {
 /// org-element-drawer-re NAME: `(any ?- ?_ word)` — hyphen, underscore,
 /// or Unicode word characters (letters and digits). `:END:` is the closer.
 /// org-element plain link at column 0: `file+emacs:` / `file+sys:` /
-/// `file:` / `shell:` / `elisp:` / `http://` / `https://` / `mailto:` /
+/// `file:` / `shell:` / `elisp:` / `help:` / `info:` / `http://` /
+/// `https://` / `mailto:` /
 /// `news:` / `doi:` / `ftp://` / `attachment:` / `id:` plus the path.
 /// Leftover after the path is hung Prose. `file+emacs:` / `file+sys:`
 /// must be matched before `file:` or the `+…` is eaten as the path.
@@ -184,6 +185,10 @@ pub(crate) fn org_plain_link_marker_len(line: &str) -> Option<usize> {
         "shell:"
     } else if t.starts_with("elisp:") {
         "elisp:"
+    } else if t.starts_with("help:") {
+        "help:"
+    } else if t.starts_with("info:") {
+        "info:"
     } else if t.starts_with("https://") {
         "https://"
     } else if t.starts_with("http://") {
diff --git a/src/sentence/unicode.rs b/src/sentence/unicode.rs
index f730033..2e7dfb7 100644
--- a/src/sentence/unicode.rs
+++ b/src/sentence/unicode.rs
@@ -3373,11 +3373,18 @@ fn split_double_bang_then_letter(segments: Vec<String>) -> Vec<String> {
         while idx + 1 < chars.len() {
             if chars[idx].1 == '!' && chars[idx + 1].1 == '!' {
                 if let Some(&(off, ch)) = chars.get(idx + 2) {
-                    if ch.is_alphabetic() && !wrap_closes_after_bang(&seg[start..], off - start) {
-                        out.push(seg[start..off].to_string());
-                        start = off;
-                        idx += 2;
-                        continue;
+                    if ch.is_alphabetic() {
+                        let prev = idx.checked_sub(1).map(|i| chars[i].1);
+                        let quote_before = matches!(
+                            prev,
+                            Some('"' | '\'' | '\u{201C}' | '\u{2018}' | '\u{00AB}')
+                        );
+                        if !quote_before && !wrap_closes_after_bang(&seg[start..], off - start) {
+                            out.push(seg[start..off].to_string());
+                            start = off;
+                            idx += 2;
+                            continue;
+                        }
                     }
                 }
             }
diff --git a/tests/org_file_token_punct.rs b/tests/org_file_token_punct.rs
index d4ff3de..484f3d5 100644
--- a/tests/org_file_token_punct.rs
+++ b/tests/org_file_token_punct.rs
@@ -130,6 +130,36 @@ fn leftover_elisp_plain_link_after_path_hangs_and_splits() {
     assert_eq!(format_text(&out, &org_cfg()).unwrap(), out);
 }
 
+#[test]
+fn leftover_help_plain_link_after_path_hangs_and_splits() {
+    let input = concat!("help:org leftover. Next.\n", "After. Next.\n",);
+    let out = format_text(input, &org_cfg()).unwrap();
+    assert!(
+        !out.contains("help:org leftover. Next."),
+        "leftover after help path must still split, got:\n{out}"
+    );
+    assert!(
+        out.contains("After.\nNext."),
+        "following prose must still split, got:\n{out}"
+    );
+    assert_eq!(format_text(&out, &org_cfg()).unwrap(), out);
+}
+
+#[test]
+fn leftover_info_plain_link_after_path_hangs_and_splits() {
+    let input = concat!("info:org leftover. Next.\n", "After. Next.\n",);
+    let out = format_text(input, &org_cfg()).unwrap();
+    assert!(
+        !out.contains("info:org leftover. Next."),
+        "leftover after info path must still split, got:\n{out}"
+    );
+    assert!(
+        out.contains("After.\nNext."),
+        "following prose must still split, got:\n{out}"
+    );
+    assert_eq!(format_text(&out, &org_cfg()).unwrap(), out);
+}
+
 #[test]
 fn leftover_attachment_plain_link_after_path_hangs_and_splits() {
     let input = concat!("attachment:plot.png leftover. Next.\n", "After. Next.\n",);
diff --git a/tests/sentence_delim_props.rs b/tests/sentence_delim_props.rs
index 2e09b2c..e93b0ad 100644
--- a/tests/sentence_delim_props.rs
+++ b/tests/sentence_delim_props.rs
@@ -100,6 +100,26 @@ fn plaintext_period_then_latex_quotes_is_idempotent() {
     );
 }
 
+#[test]
+fn plaintext_nested_quoted_bang_letter_is_span_safe() {
+    // ubuntu CI seed: `"!!a"` nested in a delimiter soup.
+    let input = "\"\"\"`A`]]\"`A`']']]\"!!a\"])]]'''}";
+    let out = format_plain(input);
+    assert_eq!(
+        format_plain(&out),
+        out,
+        "idempotence\n in={input:?}\n out={out:?}"
+    );
+    assert!(
+        newlines_respect_delimiter_spans(&out),
+        "span newline\n in={input:?}\n out={out:?}"
+    );
+    assert!(
+        !out.contains("!!\n"),
+        "must not split !!a inside quotes\n in={input:?}\n out={out:?}"
+    );
+}
+
 #[test]
 fn plaintext_quoted_bang_letter_is_span_safe() {
     let input = "\"!!a\"";