diff --git a/crates/nuwiki-lsp/src/commands.rs b/crates/nuwiki-lsp/src/commands.rs
index 3bd4bfe..3dbab64 100644
--- a/crates/nuwiki-lsp/src/commands.rs
+++ b/crates/nuwiki-lsp/src/commands.rs
@@ -3671,42 +3671,63 @@ pub mod export_ops {
         })
     }
 
-    /// Write `<html_path>/rss.xml` listing the wiki's diary entries
-    /// newest-first. The feed is intentionally minimal — just enough to
-    /// validate against RSS 2.0 parsers — since the diary URLs are file:
-    /// scheme and won't survive remote consumption anyway.
+    /// Write `<html_path>/<rss_name>` listing the wiki's diary entries
+    /// newest-first, capped by `rss_max_items`. Emits an RSS 2.0 feed with
+    /// upstream's structure: `xmlns:atom`, channel `<link>` → the diary index
+    /// page, an `<atom:link rel="self">`, channel + item `<pubDate>`, per-item
+    /// `<guid isPermaLink="false">` (bare date), and a `<![CDATA[…]]>`
+    /// rendered-HTML body per entry.
     pub fn write_rss(cfg: &WikiConfig, entries: &[DiaryEntry]) -> std::io::Result<PathBuf> {
         let mut sorted: Vec<&DiaryEntry> = entries.iter().collect();
         sorted.sort_by(|a, b| b.date.cmp(&a.date));
+        let cap = if cfg.html.rss_max_items == 0 {
+            usize::MAX
+        } else {
+            cfg.html.rss_max_items
+        };
+        let base_url = cfg.html.base_url.trim();
+        let diary_rel = cfg.diary_rel_path.trim_matches('/');
+
         let mut xml = String::from(
-            r#"<?xml version="1.0" encoding="UTF-8"?>
-<rss version="2.0"><channel>
-"#,
+            "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n\
+<rss version=\"2.0\" xmlns:atom=\"http://www.w3.org/2005/Atom\"><channel>\n",
         );
         xml.push_str(&format!("<title>{}</title>\n", xml_escape(&cfg.name)));
-        // `base_url` (per-wiki) prefixes the feed + item links so a hosted
-        // feed points at public URLs rather than local `file://` paths.
-        let base_url = cfg.html.base_url.trim();
+        // Channel link points at the diary index page (upstream), not bare base.
         if !base_url.is_empty() {
-            xml.push_str(&format!("<link>{}</link>\n", xml_escape(base_url)));
+            let index_link = format!("{base_url}{diary_rel}/{}.html", cfg.diary_index);
+            xml.push_str(&format!("<link>{}</link>\n", xml_escape(&index_link)));
+            xml.push_str(&format!(
+                "<atom:link href=\"{}{}\" rel=\"self\" type=\"application/rss+xml\" />\n",
+                xml_escape(base_url),
+                xml_escape(&cfg.html.rss_name)
+            ));
         }
         xml.push_str("<description>nuwiki diary</description>\n");
-        for e in sorted {
-            // With base_url set, build a public URL from the diary path + date;
-            // otherwise fall back to the entry's `file://` URI.
+        if let Some(first) = sorted.first() {
+            xml.push_str(&format!("<pubDate>{}</pubDate>\n", rfc822(&first.date)));
+        }
+        for e in sorted.into_iter().take(cap) {
             let link = if base_url.is_empty() {
                 e.uri.as_str().to_string()
             } else {
-                format!(
-                    "{base_url}{}/{}.html",
-                    cfg.diary_rel_path.trim_matches('/'),
-                    e.date.format()
-                )
+                format!("{base_url}{diary_rel}/{}.html", e.date.format())
             };
             xml.push_str("<item>");
             xml.push_str(&format!("<title>{}</title>", e.date.format()));
             xml.push_str(&format!("<link>{}</link>", xml_escape(&link)));
-            xml.push_str(&format!("<guid>{}</guid>", xml_escape(&link)));
+            xml.push_str(&format!(
+                "<guid isPermaLink=\"false\">{}</guid>",
+                xml_escape(&e.date.format())
+            ));
+            xml.push_str(&format!("<pubDate>{}</pubDate>", rfc822(&e.date)));
+            let body = render_entry_body(cfg, e);
+            if !body.is_empty() {
+                xml.push_str(&format!(
+                    "<description><![CDATA[{}]]></description>",
+                    cdata_safe(&body)
+                ));
+            }
             xml.push_str("</item>\n");
         }
         xml.push_str("</channel></rss>\n");
@@ -3719,8 +3740,61 @@ pub mod export_ops {
         Ok(path)
     }
 
+    /// Render a diary entry's page body (no template) for the RSS CDATA block,
+    /// honouring the wiki's colour/valid-tag/emoji HTML options. Returns an
+    /// empty string when the file can't be read.
+    fn render_entry_body(cfg: &WikiConfig, e: &DiaryEntry) -> String {
+        use nuwiki_core::render::{HtmlRenderer, Renderer};
+        use nuwiki_core::syntax::{vimwiki::VimwikiSyntax, SyntaxPlugin};
+        let Ok(path) = e.uri.to_file_path() else {
+            return String::new();
+        };
+        let Ok(src) = std::fs::read_to_string(&path) else {
+            return String::new();
+        };
+        let ast = VimwikiSyntax::new().parse(&src);
+        let mut r = HtmlRenderer::new();
+        if !cfg.html.color_dic.is_empty() {
+            r = r.with_colors(cfg.html.color_dic.clone());
+        }
+        if !cfg.html.valid_html_tags.is_empty() {
+            r = r.with_valid_html_tags(cfg.html.valid_html_tags.clone());
+        }
+        r = r.with_emoji(cfg.html.emoji_enable);
+        r.render_to_string(&ast).unwrap_or_default()
+    }
+
+    /// Make rendered HTML safe to embed inside a `<![CDATA[…]]>` block by
+    /// splitting any literal `]]>` sequence.
+    fn cdata_safe(s: &str) -> String {
+        s.replace("]]>", "]]]]><![CDATA[>")
+    }
+
+    /// RFC-822 date for a diary date at midnight UTC (1970-01-01 was a
+    /// Thursday). Used for RSS `<pubDate>`.
+    fn rfc822(d: &nuwiki_core::date::DiaryDate) -> String {
+        const WD: [&str; 7] = ["Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"];
+        const MO: [&str; 12] = [
+            "Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
+        ];
+        let days = d.to_days_epoch();
+        let wd = (((days % 7) + 4) % 7 + 7) % 7; // 1970-01-01 = Thursday (4)
+        format!(
+            "{}, {:02} {} {} 00:00:00 +0000",
+            WD[wd as usize],
+            d.day,
+            MO[(d.month.clamp(1, 12) - 1) as usize],
+            d.year
+        )
+    }
+
     fn rss_path(cfg: &WikiConfig) -> PathBuf {
-        cfg.html.html_path.join("rss.xml")
+        let name = if cfg.html.rss_name.trim().is_empty() {
+            "rss.xml"
+        } else {
+            cfg.html.rss_name.trim()
+        };
+        cfg.html.html_path.join(name)
     }
 
     fn xml_escape(s: &str) -> String {
diff --git a/crates/nuwiki-lsp/tests/html_export.rs b/crates/nuwiki-lsp/tests/html_export.rs
index 65f4d97..a25f212 100644
--- a/crates/nuwiki-lsp/tests/html_export.rs
+++ b/crates/nuwiki-lsp/tests/html_export.rs
@@ -613,13 +613,48 @@ fn write_rss_uses_base_url_for_public_links() {
     }];
     let path = nuwiki_lsp::commands::export_ops::write_rss(&c, &entries).unwrap();
     let xml = std::fs::read_to_string(&path).unwrap();
-    // Channel + item links use the public base URL, not file://.
-    assert!(xml.contains("<link>https://example.com/wiki/</link>"));
+    // Channel link points at the diary index page (upstream fidelity); item
+    // link is the public per-entry URL.
+    assert!(
+        xml.contains("<link>https://example.com/wiki/diary/diary.html</link>"),
+        "channel link: {xml}"
+    );
     assert!(xml.contains("<link>https://example.com/wiki/diary/2026-05-11.html</link>"));
-    assert!(xml.contains("<guid>https://example.com/wiki/diary/2026-05-11.html</guid>"));
+    // guid is the bare date, isPermaLink=false.
+    assert!(
+        xml.contains("<guid isPermaLink=\"false\">2026-05-11</guid>"),
+        "guid: {xml}"
+    );
+    // atom:link self + a pubDate are present.
+    assert!(xml.contains("rel=\"self\""), "atom:link: {xml}");
+    assert!(xml.contains("<pubDate>"), "pubDate: {xml}");
     assert!(!xml.contains("file:///tmp/diary"));
 }
 
+#[test]
+fn write_rss_honours_rss_name_and_max_items() {
+    use nuwiki_lsp::diary::DiaryEntry;
+    use tower_lsp::lsp_types::Url;
+    let root = temp_root("rss_cap");
+    let mut c = WikiConfig::from_root(root.clone());
+    c.html = HtmlConfig::from_root(&root);
+    c.html.rss_name = "feed.xml".into();
+    c.html.rss_max_items = 2;
+    let entries: Vec<DiaryEntry> = (1..=5)
+        .map(|d| DiaryEntry {
+            date: DiaryDate::from_ymd(2026, 5, d).unwrap(),
+            uri: Url::parse(&format!("file:///tmp/diary/2026-05-0{d}.wiki")).unwrap(),
+        })
+        .collect();
+    let path = nuwiki_lsp::commands::export_ops::write_rss(&c, &entries).unwrap();
+    assert!(path.ends_with("feed.xml"), "rss_name: {path:?}");
+    let xml = std::fs::read_to_string(&path).unwrap();
+    // Capped at 2 items (newest first: 05-05, 05-04).
+    assert_eq!(xml.matches("<item>").count(), 2, "cap: {xml}");
+    assert!(xml.contains("2026-05-05") && xml.contains("2026-05-04"));
+    assert!(!xml.contains("2026-05-03"), "older items dropped: {xml}");
+}
+
 #[test]
 fn write_page_invokes_custom_wiki2html_converter() {
     let root = temp_root("custom_w2h");