chore(release): 1.5.11 — cross-origin iframe scroll/click + open auto-reattach (#35, #36)

#35: `open` auto-reattaches when the bound relay tab is gone — drops the dead
page, opens a fresh tab in the session's group, and navigates it, instead of
only `tab new` recovering.

#36: scroll and click now reach content inside cross-origin OOPIFs:
- scroll dispatches a real wheel at a viewport point (default center, --at x,y,
  or --frame n) so it scrolls the iframe under the pointer, which
  window.scrollBy on the top document silently no-ops on.
- over the extension relay, clicks always use DOM-dispatch instead of
  coordinate Input events — a coordinate event isn't confined to the target tab
  on a busy real Chrome (it drifted onto the foreground tab) and an OOPIF
  element's box can't be mapped to a top-viewport point.

cli/Cargo.lock

@@ -290,7 +290,7 @@ checksum = "613afe47fcd5fac7ccf1db93babcb082c5994d996f20b8b159f2ad1658eb5724"
 
 [[package]]
 name = "chrome-use"
-version = "1.5.8"
+version = "1.5.11"
 dependencies = [
  "aes",
  "aes-gcm",

cli/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "chrome-use"
-version = "1.5.8"
+version = "1.5.11"
 edition = "2021"
 description = "Fast browser automation CLI for AI agents"
 license = "Apache-2.0"

cli/src/commands.rs

@@ -729,10 +729,56 @@ fn parse_command_inner(args: &[String], flags: &Flags) -> Result<Value, ParseErr
                         } else {
                             return Err(ParseError::MissingArguments {
                                 context: "scroll --selector".to_string(),
-                                usage: "scroll [direction] [amount] [--selector <sel>]",
+                                usage: "scroll [direction] [amount] [--selector <sel>] [--at <x,y>] [--frame <n>]",
                             });
                         }
                     }
+                    "--at" => {
+                        // `--at x,y`: dispatch the wheel at this viewport pixel, so it
+                        // scrolls whatever element/iframe is under the pointer — including
+                        // cross-origin iframes that `window.scrollBy` can't reach (#36).
+                        let val = rest.get(i + 1).ok_or(ParseError::MissingArguments {
+                            context: "scroll --at".to_string(),
+                            usage: "scroll [direction] [amount] --at <x,y>",
+                        })?;
+                        let mut parts = val.split(',');
+                        match (
+                            parts.next().and_then(|s| s.trim().parse::<f64>().ok()),
+                            parts.next().and_then(|s| s.trim().parse::<f64>().ok()),
+                        ) {
+                            (Some(x), Some(y)) => {
+                                obj.insert("at".to_string(), json!([x, y]));
+                            }
+                            _ => {
+                                return Err(ParseError::InvalidValue {
+                                    message: format!("scroll --at: invalid coordinate `{}`", val),
+                                    usage: "scroll [direction] [amount] --at <x,y> (e.g. --at 640,400)",
+                                })
+                            }
+                        }
+                        i += 1;
+                    }
+                    "--frame" => {
+                        // `--frame n`: scroll the n-th frame from `chrome-use frames` by
+                        // dispatching the wheel at that frame's center — reaches content in
+                        // a cross-origin iframe without needing a selector into it (#36).
+                        let val = rest.get(i + 1).ok_or(ParseError::MissingArguments {
+                            context: "scroll --frame".to_string(),
+                            usage: "scroll [direction] [amount] --frame <n>",
+                        })?;
+                        match val.trim().parse::<usize>() {
+                            Ok(n) => {
+                                obj.insert("frame".to_string(), json!(n));
+                            }
+                            Err(_) => {
+                                return Err(ParseError::InvalidValue {
+                                    message: format!("scroll --frame: invalid index `{}`", val),
+                                    usage: "scroll [direction] [amount] --frame <n> (index from `chrome-use frames`)",
+                                })
+                            }
+                        }
+                        i += 1;
+                    }
                     arg if arg.starts_with('-') => {}
                     _ => {
                         match positional_index {
@@ -907,17 +953,37 @@ fn parse_command_inner(args: &[String], flags: &Flags) -> Result<Value, ParseErr
             // selector: @ref or CSS selector
             // path: file path (contains / or . or ends with known extension)
             let mut full_page = false;
-            let positional: Vec<&str> = rest
+            let mut clip: Option<Value> = None;
-                .iter()
+            let mut positional: Vec<&str> = Vec::new();
-                .filter(|arg| match **arg {
+            let mut i = 0;
-                    "--full" | "-f" => {
+            while i < rest.len() {
-                        full_page = true;
+                match rest[i] {
-                        false
+                    "--full" | "-f" => full_page = true,
+                    // `--clip x,y,w,h` captures a pixel region (issue #34).
+                    "--clip" => {
+                        let raw = rest.get(i + 1).ok_or_else(|| ParseError::MissingArguments {
+                            context: "screenshot --clip".to_string(),
+                            usage: "screenshot --clip <x,y,w,h> [path]",
+                        })?;
+                        let nums: Vec<f64> = raw
+                            .split(',')
+                            .filter_map(|n| n.trim().parse::<f64>().ok())
+                            .collect();
+                        if nums.len() != 4 {
+                            return Err(ParseError::InvalidValue {
+                                message: format!("--clip expects 'x,y,w,h' (4 numbers), got '{raw}'"),
+                                usage: "screenshot --clip <x,y,w,h> [path]",
+                            });
+                        }
+                        clip = Some(json!({
+                            "x": nums[0], "y": nums[1], "width": nums[2], "height": nums[3]
+                        }));
+                        i += 1;
                     }
-                    _ => true,
+                    other => positional.push(other),
-                })
+                }
-                .copied()
+                i += 1;
-                .collect();
+            }
             let (selector, path) = match (positional.first(), positional.get(1)) {
                 (Some(first), Some(second)) => {
                     // Two args: first is selector, second is path
@@ -948,6 +1014,9 @@ fn parse_command_inner(args: &[String], flags: &Flags) -> Result<Value, ParseErr
                 "path": path, "selector": selector,
                 "fullPage": full_page, "annotate": flags.annotate
             });
+            if let Some(c) = clip {
+                cmd["clip"] = c;
+            }
             if let Some(ref fmt) = flags.screenshot_format {
                 cmd["format"] = json!(fmt);
             }
@@ -4327,6 +4396,21 @@ mod tests {
         assert_eq!(cmd["fullPage"], true);
     }
 
+    #[test]
+    fn test_screenshot_clip() {
+        // `--clip x,y,w,h` captures a pixel region (issue #34); the path still parses.
+        let cmd = parse_command(&args("screenshot --clip 10,20,200,40 out.png"), &default_flags())
+            .unwrap();
+        assert_eq!(cmd["action"], "screenshot");
+        assert_eq!(cmd["clip"]["x"], 10.0);
+        assert_eq!(cmd["clip"]["y"], 20.0);
+        assert_eq!(cmd["clip"]["width"], 200.0);
+        assert_eq!(cmd["clip"]["height"], 40.0);
+        assert_eq!(cmd["path"], "out.png");
+        // Bad clip is a clear error, not silent.
+        assert!(parse_command(&args("screenshot --clip 1,2,3"), &default_flags()).is_err());
+    }
+
     #[test]
     fn test_screenshot_with_ref() {
         let cmd = parse_command(&args("screenshot @e1"), &default_flags()).unwrap();
@@ -5830,6 +5914,35 @@ mod tests {
         assert_eq!(cmd["selector"], ".sidebar");
     }
 
+    #[test]
+    fn test_scroll_at_coordinate() {
+        // `--at x,y` carries a [x, y] array for a wheel dispatched at that pixel
+        // (issue #36: cross-origin iframe scroll).
+        let cmd = parse_command(&args("scroll down 700 --at 640,400"), &default_flags()).unwrap();
+        assert_eq!(cmd["action"], "scroll");
+        assert_eq!(cmd["direction"], "down");
+        assert_eq!(cmd["amount"], 700);
+        assert_eq!(cmd["at"], json!([640.0, 400.0]));
+    }
+
+    #[test]
+    fn test_scroll_at_rejects_garbage() {
+        assert!(parse_command(&args("scroll --at nope"), &default_flags()).is_err());
+        assert!(parse_command(&args("scroll --at 1"), &default_flags()).is_err());
+    }
+
+    #[test]
+    fn test_scroll_frame_index() {
+        let cmd = parse_command(&args("scroll down 700 --frame 2"), &default_flags()).unwrap();
+        assert_eq!(cmd["action"], "scroll");
+        assert_eq!(cmd["frame"], 2);
+    }
+
+    #[test]
+    fn test_scroll_frame_rejects_non_integer() {
+        assert!(parse_command(&args("scroll --frame two"), &default_flags()).is_err());
+    }
+
     #[test]
     fn test_scroll_selector_before_positional() {
         let cmd =

cli/src/main.rs

@@ -1369,6 +1369,20 @@ fn main() {
         && flags.provider.is_none()
         && (flags.force_launch || !flags.auto_connect)
     {
+        // Launching a debug-port Chrome pops Chrome's "Allow remote debugging?"
+        // consent modal (Chrome 136+). When the ab-connect relay is already up,
+        // this is almost always unintended — the relay drives the user's real
+        // Chrome with NO modal. Warn so the modal is self-explained and the
+        // caller (often a stray --launch / --no-auto-connect) is fixable (#32).
+        if !flags.json && connect::relay_url().is_some() {
+            eprintln!(
+                "{} launching a new Chrome with a debug port — this pops Chrome's \
+                 \"Allow remote debugging?\" modal.\n  The ab-connect relay is up; \
+                 drop --launch/--new (and don't pass --no-auto-connect) to drive your \
+                 real Chrome with no modal.",
+                color::warning_indicator()
+            );
+        }
         let mut launch_cmd = json!({
             "id": gen_id(),
             "action": "launch",

cli/src/native/actions.rs

@@ -3000,6 +3000,14 @@ async fn handle_screenshot(cmd: &Value, state: &mut DaemonState) -> Result<Value
             .get("screenshotDir")
             .and_then(|v| v.as_str())
             .map(String::from),
+        clip: cmd.get("clip").and_then(|c| {
+            Some((
+                c.get("x")?.as_f64()?,
+                c.get("y")?.as_f64()?,
+                c.get("width")?.as_f64()?,
+                c.get("height")?.as_f64()?,
+            ))
+        }),
     };
 
     if annotate {
@@ -3459,17 +3467,169 @@ async fn handle_scroll(cmd: &Value, state: &mut DaemonState) -> Result<Value, St
         }
     }
 
-    interaction::scroll(
+    // An explicit `--selector` keeps the precise element-scroll path (scrollBy on
-        &mgr.client,
+    // the resolved node, same-origin only).
-        &session_id,
+    if let Some(sel) = selector {
-        &state.ref_map,
+        interaction::scroll(
-        selector,
+            &mgr.client,
-        dx,
+            &session_id,
-        dy,
+            &state.ref_map,
-        &state.iframe_sessions,
+            Some(sel),
-    )
+            dx,
-    .await?;
+            dy,
-    Ok(json!({ "scrolled": true }))
+            &state.iframe_sessions,
+        )
+        .await?;
+        return Ok(json!({ "scrolled": true, "via": "selector" }));
+    }
+
+    // No selector: dispatch a real (isTrusted) wheel at a viewport coordinate.
+    // This hits the compositor and scrolls whatever scroll container is under the
+    // pointer — including cross-origin iframes that `window.scrollBy` on the top
+    // document silently no-ops on (issue #36). The coordinate is, in priority:
+    //   --at x,y  → that exact pixel
+    //   --frame n → the center of frame n from `chrome-use frames`
+    //   default   → the viewport center
+    let (x, y, via) = if let Some(at) = cmd.get("at").and_then(|v| v.as_array()) {
+        let x = at.first().and_then(|v| v.as_f64()).unwrap_or(0.0);
+        let y = at.get(1).and_then(|v| v.as_f64()).unwrap_or(0.0);
+        (x, y, "at")
+    } else if let Some(n) = cmd.get("frame").and_then(|v| v.as_u64()) {
+        let (x, y) = frame_center(mgr, &session_id, &state.iframe_sessions, n as usize).await?;
+        (x, y, "frame")
+    } else {
+        let (x, y) = viewport_center(mgr, &session_id).await?;
+        (x, y, "center")
+    };
+
+    dispatch_wheel(&mgr.client, &session_id, x, y, dx, dy).await?;
+    Ok(json!({ "scrolled": true, "via": via, "at": [x, y] }))
+}
+
+/// Viewport center in CSS pixels, used as the default wheel landing point for
+/// `scroll` (issue #36). Falls back to a sane 640×400 center if the page can't
+/// be evaluated (e.g. a restricted document).
+async fn viewport_center(mgr: &BrowserManager, session_id: &str) -> Result<(f64, f64), String> {
+    let dims = mgr
+        .client
+        .send_command_typed::<_, Value>(
+            "Runtime.evaluate",
+            &super::cdp::types::EvaluateParams {
+                expression: "[window.innerWidth, window.innerHeight]".to_string(),
+                return_by_value: Some(true),
+                await_promise: Some(false),
+            },
+            Some(session_id),
+        )
+        .await
+        .ok();
+    let arr = dims
+        .as_ref()
+        .and_then(|v| v.get("result"))
+        .and_then(|v| v.get("value"))
+        .and_then(|v| v.as_array());
+    let w = arr
+        .and_then(|a| a.first())
+        .and_then(|v| v.as_f64())
+        .filter(|w| *w > 0.0)
+        .unwrap_or(1280.0);
+    let h = arr
+        .and_then(|a| a.get(1))
+        .and_then(|v| v.as_f64())
+        .filter(|h| *h > 0.0)
+        .unwrap_or(800.0);
+    Ok((w / 2.0, h / 2.0))
+}
+
+/// Center of the `n`-th frame (as listed by `chrome-use frames`) in top-viewport
+/// CSS pixels, so `scroll --frame n` lands its wheel inside a cross-origin iframe
+/// without needing a selector into it (issue #36). Resolves the frame's owning
+/// `<iframe>` element box via `DOM.getFrameOwner` + `DOM.getBoxModel` — exact for
+/// a frame nested directly under the top document; for a deeper nesting the box is
+/// relative to the intermediate frame, so prefer `--at x,y` from a screenshot.
+async fn frame_center(
+    mgr: &BrowserManager,
+    session_id: &str,
+    iframe_sessions: &HashMap<String, String>,
+    n: usize,
+) -> Result<(f64, f64), String> {
+    let frames =
+        super::element::collect_all_frames_text(&mgr.client, session_id, iframe_sessions).await?;
+    let frame = frames.get(n).ok_or_else(|| {
+        format!(
+            "frame index {} out of range (run `chrome-use frames`: {} frame(s))",
+            n,
+            frames.len()
+        )
+    })?;
+    if n == 0 {
+        // Frame 0 is the top document — there's no owner element; scroll its center.
+        return viewport_center(mgr, session_id).await;
+    }
+    let owner = mgr
+        .client
+        .send_command_typed::<_, Value>(
+            "DOM.getFrameOwner",
+            &json!({ "frameId": frame.frame_id }),
+            Some(session_id),
+        )
+        .await
+        .map_err(|e| format!("can't locate frame {}'s owner element: {}", n, e))?;
+    let backend_node_id = owner
+        .get("backendNodeId")
+        .and_then(|v| v.as_i64())
+        .ok_or_else(|| format!("frame {} has no owner <iframe> element", n))?;
+    let box_model = mgr
+        .client
+        .send_command_typed::<_, Value>(
+            "DOM.getBoxModel",
+            &json!({ "backendNodeId": backend_node_id }),
+            Some(session_id),
+        )
+        .await
+        .map_err(|e| format!("can't measure frame {}'s box: {}", n, e))?;
+    let content = box_model
+        .get("model")
+        .and_then(|m| m.get("content"))
+        .and_then(|c| c.as_array())
+        .ok_or_else(|| format!("frame {} box model has no content quad", n))?;
+    let coord = |i: usize| content.get(i).and_then(|v| v.as_f64()).unwrap_or(0.0);
+    // content quad is [x1,y1, x2,y2, x3,y3, x4,y4]; opposite corners are 0 and 2.
+    let cx = (coord(0) + coord(4)) / 2.0;
+    let cy = (coord(1) + coord(5)) / 2.0;
+    Ok((cx, cy))
+}
+
+/// Dispatch a trusted mouse wheel at `(x, y)`, humanized like `handle_wheel`.
+async fn dispatch_wheel(
+    client: &super::cdp::client::CdpClient,
+    session_id: &str,
+    x: f64,
+    y: f64,
+    delta_x: f64,
+    delta_y: f64,
+) -> Result<(), String> {
+    let level = humanize::active_level();
+    let seed = humanize::next_seed();
+    for (dx, dy, delay) in humanize::scroll_segments(delta_x, delta_y, level, seed) {
+        client
+            .send_command(
+                "Input.dispatchMouseEvent",
+                Some(json!({
+                    "type": "mouseWheel",
+                    "x": x,
+                    "y": y,
+                    "deltaX": dx,
+                    "deltaY": dy,
+                })),
+                Some(session_id),
+            )
+            .await?;
+        if !delay.is_zero() {
+            tokio::time::sleep(delay).await;
+        }
+    }
+    Ok(())
 }
 
 async fn handle_select(cmd: &Value, state: &mut DaemonState) -> Result<Value, String> {
@@ -7446,6 +7606,7 @@ async fn handle_diff_screenshot(cmd: &Value, state: &DaemonState) -> Result<Valu
         quality: None,
         annotate: false,
         output_dir: None,
+        clip: None,
     };
 
     let result = screenshot::take_screenshot(

cli/src/native/browser.rs

@@ -121,7 +121,7 @@ fn normalize_url_for_match(url: &str) -> String {
 fn update_page_target_info_in_pages(pages: &mut [PageInfo], target: &TargetInfo) -> bool {
     if let Some(page) = pages.iter_mut().find(|p| p.target_id == target.target_id) {
         page.url = target.url.clone();
-        page.title = target.title.clone();
+        page.title = sanitize_title(&target.title);
         page.target_type = target.target_type.clone();
         return true;
     }
@@ -166,6 +166,29 @@ fn resolve_active_index(
     active_page_index
 }
 
+/// Strip zero-width / invisible / bidi-format Unicode from a page title before
+/// we store it. Some sites prepend runs of ZWJ / word-joiner / invisible-times /
+/// BOM to `document.title` (badging, watermarking, anti-scrape); left in, they
+/// pollute `tab list`, break text matching, and wreck column alignment (#33).
+fn sanitize_title(s: &str) -> String {
+    s.chars()
+        .filter(|&c| {
+            !matches!(c as u32,
+                0x00AD            // soft hyphen
+                | 0x200B..=0x200F // ZWSP, ZWNJ, ZWJ, LRM, RLM
+                | 0x2028 | 0x2029 // line / paragraph separators
+                | 0x202A..=0x202E // bidi embedding/override
+                | 0x2060..=0x2064 // word joiner, invisible operators
+                | 0x2066..=0x2069 // bidi isolates
+                | 0x180E          // Mongolian vowel separator
+                | 0xFEFF          // BOM / ZW no-break space
+            )
+        })
+        .collect::<String>()
+        .trim()
+        .to_string()
+}
+
 /// Best-effort MIME type from a filename extension, for the relay file-upload
 /// fallback (the page-constructed `File` needs a sensible `type`). Covers the
 /// common upload kinds; anything unknown falls back to a generic binary type.
@@ -231,6 +254,21 @@ fn active_index_is_owned(
         .unwrap_or(false)
 }
 
+/// Whether a CDP error means the bound relay target is gone — the tab was
+/// closed, navigated across processes (renderer swap), or lost after an
+/// extension/service-worker restart, and the relay could not re-attach. The
+/// ab-connect relay surfaces these as `stale sessionId … its tab is gone`,
+/// `unknown sessionId …`, or `no attached tab …`. `navigate` keys its
+/// auto-reattach recovery off this (issue #35) so a dead session rebinds to a
+/// fresh tab instead of erroring on every command until the user runs `tab new`.
+fn is_stale_target_error(error: &str) -> bool {
+    let lower = error.to_lowercase();
+    lower.contains("its tab is gone")
+        || lower.contains("stale sessionid")
+        || lower.contains("unknown sessionid")
+        || lower.contains("no attached tab")
+}
+
 /// Converts common error messages into AI-friendly, actionable descriptions.
 pub fn to_ai_friendly_error(error: &str) -> String {
     let lower = error.to_lowercase();
@@ -773,7 +811,7 @@ impl BrowserManager {
                     target_id: target.target_id.clone(),
                     session_id: attach_result.session_id.clone(),
                     url: target.url.clone(),
-                    title: target.title.clone(),
+                    title: sanitize_title(&target.title),
                     target_type: target.target_type.clone(),
                 });
             }
@@ -892,6 +930,24 @@ impl BrowserManager {
         )
     }
 
+    /// Drop the page bound to `session_id` from the tracked list — used when the
+    /// relay reports its tab is gone (issue #35) so the stale entry can't keep
+    /// resolving as active. Forgets ownership, unpins it if it was pinned, and
+    /// keeps `active_page_index` in range.
+    fn drop_page_by_session(&mut self, session_id: &str) {
+        let Some(pos) = self.pages.iter().position(|p| p.session_id == session_id) else {
+            return;
+        };
+        let target_id = self.pages[pos].target_id.clone();
+        self.pages.remove(pos);
+        self.created_targets.remove(&target_id);
+        if self.active_target_id.as_deref() == Some(target_id.as_str()) {
+            self.active_target_id = None;
+        }
+        self.active_page_index =
+            active_page_index_after_removal(self.active_page_index, pos, self.pages.len());
+    }
+
     /// Pin the current active page by target_id so later commands stick to it.
     /// Call after any explicit open / tab new / tab switch.
     fn pin_active_target(&mut self) {
@@ -921,10 +977,10 @@ impl BrowserManager {
         if self.agent_group().is_some() && !self.active_is_session_owned() {
             self.tab_new(None, None).await?;
         }
-        let session_id = self.active_session_id()?.to_string();
+        let mut session_id = self.active_session_id()?.to_string();
         let mut lifecycle_rx = self.client.subscribe();
 
-        let nav_result: PageNavigateResult = self
+        let nav_result: PageNavigateResult = match self
             .client
             .send_command_typed(
                 "Page.navigate",
@@ -934,7 +990,38 @@ impl BrowserManager {
                 },
                 Some(&session_id),
             )
-            .await?;
+            .await
+        {
+            Ok(r) => r,
+            // Auto-reattach when the bound tab is gone (issue #35). On the shared
+            // real browser the human can close/swap the agent's tab, and a
+            // cross-process nav can destroy the target without a re-attachable
+            // tabId — both leave the cached `cb-tab-<id>` session stale, so every
+            // command (including `open`) failed on it and only `tab new`
+            // recovered. The relay error literally says "re-open your target URL
+            // to re-attach"; fulfil that here: drop the dead page, open a fresh
+            // owned tab in this session's group, and navigate THAT. Gated on the
+            // relay (`agent_group`) and on the explicit navigation intent — read
+            // commands deliberately still fail loudly rather than silently
+            // recover onto a blank tab and return wrong data (issue #8.1).
+            Err(e) if self.agent_group().is_some() && is_stale_target_error(&e) => {
+                self.drop_page_by_session(&session_id);
+                self.tab_new(None, None).await?;
+                session_id = self.active_session_id()?.to_string();
+                lifecycle_rx = self.client.subscribe();
+                self.client
+                    .send_command_typed(
+                        "Page.navigate",
+                        &PageNavigateParams {
+                            url: url.to_string(),
+                            referrer: None,
+                        },
+                        Some(&session_id),
+                    )
+                    .await?
+            }
+            Err(e) => return Err(e),
+        };
 
         if let Some(ref error_text) = nav_result.error_text {
             return Err(format!("Navigation failed: {}", error_text));
@@ -999,7 +1086,7 @@ impl BrowserManager {
         self.active_page_index = self.resolved_active_index();
         if let Some(page) = self.pages.get_mut(self.active_page_index) {
             page.url = page_url.clone();
-            page.title = title.clone();
+            page.title = sanitize_title(&title);
         }
         self.pin_active_target();
 
@@ -1061,7 +1148,7 @@ impl BrowserManager {
 
     pub async fn get_title(&self) -> Result<String, String> {
         let result = self.evaluate_simple("document.title").await?;
-        Ok(result.as_str().unwrap_or("").to_string())
+        Ok(sanitize_title(result.as_str().unwrap_or("")))
     }
 
     pub async fn get_content(&self) -> Result<String, String> {
@@ -1404,7 +1491,7 @@ impl BrowserManager {
                 target_id: target.target_id.clone(),
                 session_id: attach.session_id.clone(),
                 url: target.url.clone(),
-                title: target.title.clone(),
+                title: sanitize_title(&target.title),
                 target_type: target.target_type.clone(),
             };
             self.add_background_page(page.clone());
@@ -1465,7 +1552,7 @@ impl BrowserManager {
                 target_id: target.target_id.clone(),
                 session_id: attach_result.session_id.clone(),
                 url: target.url.clone(),
-                title: target.title.clone(),
+                title: sanitize_title(&target.title),
                 target_type: target.target_type.clone(),
             });
             let _ = self.enable_domains(&attach_result.session_id).await;
@@ -1508,7 +1595,7 @@ impl BrowserManager {
                             }
                         }
                         if let Some(t) = ti.get("title").and_then(|v| v.as_str()) {
-                            page.title = t.to_string();
+                            page.title = sanitize_title(t);
                         }
                     }
                 }
@@ -1696,7 +1783,7 @@ impl BrowserManager {
 
         if let Some(page) = self.pages.get_mut(index) {
             page.url = url.clone();
-            page.title = title.clone();
+            page.title = sanitize_title(&title);
         }
 
         let page = &self.pages[index];
@@ -2670,6 +2757,27 @@ mod tests {
         assert_eq!(active_page_index_after_removal(0, 0, 0), 0);
     }
 
+    #[test]
+    fn stale_target_error_matches_relay_signatures() {
+        // The exact relay error `open` must recover from (issue #35), as wrapped
+        // by send_command's `CDP error (Page.navigate): …` prefix.
+        assert!(is_stale_target_error(
+            "CDP error (Page.navigate): stale sessionId cb-tab-1655244623 for Page.navigate: \
+             its tab is gone (closed, navigated across processes, or lost after an extension \
+             restart). Re-attach by re-opening your target URL before retrying."
+        ));
+        assert!(is_stale_target_error("unknown sessionId cb-tab-7 for Page.navigate"));
+        assert!(is_stale_target_error("no attached tab for Page.navigate"));
+    }
+
+    #[test]
+    fn stale_target_error_ignores_unrelated_failures() {
+        // A genuine navigation failure (bad URL, DNS, blocked) must NOT trigger
+        // the open-a-fresh-tab recovery — that would mask the real error.
+        assert!(!is_stale_target_error("Navigation failed: net::ERR_NAME_NOT_RESOLVED"));
+        assert!(!is_stale_target_error("CDP command timed out: Page.navigate"));
+    }
+
     fn page(target_id: &str) -> PageInfo {
         PageInfo {
             tab_id: 1,
@@ -2767,6 +2875,19 @@ mod tests {
         assert!(!active_index_is_owned(&[], None, 0, &created));
     }
 
+    #[test]
+    fn test_sanitize_title() {
+        // The exact pollution from #33: ZWJ / word-joiner / invisible-times / BOM
+        // prepended to "GitHub".
+        let dirty = "\u{200d}\u{2061}\u{200d}\u{2063}\u{200b}\u{2062}\u{feff}GitHub";
+        assert_eq!(sanitize_title(dirty), "GitHub");
+        // Clean titles (incl. CJK + normal punctuation) pass through untouched.
+        assert_eq!(sanitize_title("購入手続きへ - メルカリ"), "購入手続きへ - メルカリ");
+        assert_eq!(sanitize_title("  Hello World  "), "Hello World");
+        // Emoji and real content survive; only the invisibles are dropped.
+        assert_eq!(sanitize_title("✓ Done\u{200b}"), "✓ Done");
+    }
+
     #[test]
     fn test_mime_for_path() {
         assert_eq!(mime_for_path("a.png"), "image/png");

cli/src/native/interaction.rs

@@ -45,6 +45,31 @@ pub async fn click(
         .await;
     }
 
+    // Over the extension relay we drive the user's real, in-use Chrome, where a
+    // coordinate `Input.dispatchMouseEvent` is NOT reliably confined to our target
+    // tab — it can be delivered to whatever tab is in the foreground, and an OOPIF
+    // element's box can't be mapped to a top-viewport point at all. This twice
+    // opened an unrelated tab on the user's busy Chrome (issues #31/#36). So on the
+    // relay, never use coordinates for a normal left click: DOM-dispatch invokes
+    // the element's click in its own (frame) session, always hitting the right
+    // element in the right tab. Double/right clicks still need true pointer
+    // semantics, and `coord` mode is an explicit opt-out.
+    if mode != "coord" && button == "left" && click_count == 1 {
+        let in_iframe = parse_ref(selector_or_ref)
+            .and_then(|r| ref_map.get(&r).map(|e| e.frame_id.is_some()))
+            .unwrap_or(false);
+        if in_iframe || crate::connect::relay_url().is_some() {
+            return dom_click(
+                client,
+                session_id,
+                ref_map,
+                selector_or_ref,
+                iframe_sessions,
+            )
+            .await;
+        }
+    }
+
     let resolved = resolve_element_center(
         client,
         session_id,

cli/src/native/screenshot.rs

@@ -60,6 +60,9 @@ pub struct ScreenshotOptions {
     pub quality: Option<i32>,
     pub annotate: bool,
     pub output_dir: Option<String>,
+    /// Explicit pixel region (x, y, width, height) — `--clip` (issue #34). Takes
+    /// precedence over selector/full_page.
+    pub clip: Option<(f64, f64, f64, f64)>,
 }
 
 impl Default for ScreenshotOptions {
@@ -72,6 +75,7 @@ impl Default for ScreenshotOptions {
             quality: None,
             annotate: false,
             output_dir: None,
+            clip: None,
         }
     }
 }
@@ -187,7 +191,16 @@ async fn capture_screenshot_base64(
         capture_beyond_viewport: if options.full_page { Some(true) } else { None },
     };
 
-    if options.full_page {
+    if let Some((x, y, width, height)) = options.clip {
+        // Explicit pixel region wins over selector/full_page (issue #34).
+        params.clip = Some(Viewport {
+            x,
+            y,
+            width,
+            height,
+            scale: 1.0,
+        });
+    } else if options.full_page {
         let metrics: Value = client
             .send_command_no_params("Page.getLayoutMetrics", Some(session_id))
             .await?;

cli/src/output.rs

@@ -1735,12 +1735,23 @@ Usage: chrome-use scroll [direction] [amount] [options]
 
 Scrolls the page or a specific element in the specified direction.
 
+Without --selector, scroll dispatches a real (isTrusted) mouse wheel at a
+viewport coordinate, so it scrolls whatever container is under the pointer —
+including cross-origin iframes (Google Payments, Stripe, embedded checkout/KYC)
+that plain page scroll can't reach.
+
 Arguments:
   direction            up, down, left, right (default: down)
   amount               Pixels to scroll (default: 300)
 
 Options:
-  -s, --selector <sel> CSS selector for a scrollable container
+  -s, --selector <sel> CSS selector for a scrollable container (same-origin)
+  --at <x,y>           Dispatch the wheel at this viewport pixel (read it from a
+                       screenshot) — precise way into a cross-origin iframe
+  --frame <n>          Scroll the n-th frame from `chrome-use frames` (wheel at
+                       that frame's center)
+
+Without --selector/--at/--frame the wheel lands at the viewport center.
 
 Global Options:
   --json               Output as JSON
@@ -1752,6 +1763,8 @@ Examples:
   chrome-use scroll up 200
   chrome-use scroll left 100
   chrome-use scroll down 500 --selector "div.scroll-container"
+  chrome-use scroll down 700 --at 640,400      # wheel at a pixel over an iframe
+  chrome-use scroll down 700 --frame 2         # scroll frame 2 from `frames`
 "##
         }
         "scrollintoview" | "scrollinto" => {
@@ -1832,6 +1845,8 @@ Pass --hide-scrollbars false when launching to keep native scrollbars visible.
 
 Options:
   --full, -f           Capture full page (not just viewport)
+  [selector]           Capture just an element (CSS or @ref), e.g. `screenshot ".header" h.png`
+  --clip <x,y,w,h>     Capture a pixel region, e.g. `screenshot --clip 0,0,200,40 corner.png`
   --annotate           Overlay numbered labels on interactive elements.
                        Each label [N] corresponds to ref @eN from snapshot.
                        Prints a legend mapping labels to element roles/names.
@@ -1852,6 +1867,8 @@ Examples:
   chrome-use screenshot
   chrome-use screenshot ./screenshot.png
   chrome-use screenshot --full ./full-page.png
+  chrome-use screenshot ".header .indicator" corner.png  # just one element
+  chrome-use screenshot --clip 1600,0,200,40 corner.png  # a pixel region
   chrome-use screenshot --annotate              # Labeled screenshot + legend
   chrome-use screenshot --annotate ./page.png   # Save annotated screenshot
   chrome-use screenshot --annotate --json       # JSON output with annotations

package.json

@@ -1,6 +1,6 @@
 {
   "name": "chrome-use",
-  "version": "1.5.8",
+  "version": "1.5.11",
   "description": "chrome-use — drive your real, logged-in Chrome from any AI agent, stealth by default",
   "type": "module",
   "packageManager": "pnpm@11.1.3",

skill-data/core/SKILL.md

@@ -287,6 +287,10 @@ chrome-use upload @e5 file1.pdf        # upload file(s) — works over the exten
                                           # a File there (chunked under native-messaging's 1 MiB cap).
                                           # Works on file <input>s and drop/paste composers (e.g. X).
 chrome-use scroll down 500             # scroll page (up/down/left/right)
+chrome-use scroll down 700 --at 640,400 # wheel at a pixel — scrolls a cross-origin
+                                          # iframe (Payments/Stripe/checkout/KYC) that
+                                          # plain page scroll can't reach
+chrome-use scroll down 700 --frame 2    # scroll frame 2 from `chrome-use frames`
 chrome-use scrollintoview @e1          # scroll element into view
 chrome-use drag @e1 @e2                # drag and drop
 ```