Merge pull request '修复 Web 预检浏览器失败恢复' (#591) from fix/web-preflight-browser-recovery into master

Reviewed-on: #591

apps/ai-game-creator-shell/src-tauri/Cargo.toml

@@ -86,7 +86,7 @@ maud = "0.27.0"
 libc = "0.2"
 
 [target.'cfg(windows)'.dependencies]
-windows-sys = { version = "0.61", features = ["Wdk_Storage_FileSystem", "Win32_Foundation", "Win32_Storage_FileSystem", "Win32_System_Diagnostics_ToolHelp", "Win32_System_IO", "Win32_System_JobObjects", "Win32_System_Threading", "Win32_UI_WindowsAndMessaging"] }
+windows-sys = { version = "0.61", features = ["Wdk_Storage_FileSystem", "Win32_Foundation", "Win32_Storage_FileSystem", "Win32_System_Diagnostics_ToolHelp", "Win32_System_IO", "Win32_System_JobObjects", "Win32_System_Threading", "Win32_UI_Shell", "Win32_UI_WindowsAndMessaging"] }
 
 [profile.dev]
 opt-level = 0

apps/ai-game-creator-shell/src-tauri/src/browser/sweep.rs

@@ -12,6 +12,8 @@ use serde::{Deserialize, Serialize};
 
 use super::discovery::system_browser_candidates;
 use super::process::{browser_process_temp_root, BROWSER_TEMP_PREFIX};
+#[cfg(windows)]
+use crate::process_session::WindowsProcessJob;
 
 const OWNER_FILE: &str = "owner.json";
 const OWNER_SCHEMA: &str = "agc-browser-process.v1";
@@ -34,13 +36,17 @@ pub(super) struct BrowserProcessOwner {
 
 /// 运行期 launch 的身份锚点；任何一步拿不到身份证明都不写，
 /// 该目录之后按旧残留只删不杀。
-pub(super) fn write_browser_process_owner(temp_root: &Path, browser_pid: u32, executable: &Path) {
+pub(super) fn write_browser_process_owner(
+    temp_root: &Path,
+    browser_pid: u32,
+    executable: &Path,
+) -> bool {
     let identity =
         crate::process_identity::external_agent_runner_process_start_identity(browser_pid);
     let owner_identity =
         crate::process_identity::external_agent_runner_process_start_identity(std::process::id());
     let (Ok(Some(identity)), Ok(Some(owner_identity))) = (identity, owner_identity) else {
-        return;
+        return false;
     };
     let owner = BrowserProcessOwner {
         schema_version: OWNER_SCHEMA.into(),
@@ -52,9 +58,9 @@ pub(super) fn write_browser_process_owner(temp_root: &Path, browser_pid: u32, ex
         created_unix_ms: super::evidence::unix_time_ms(),
     };
     let Ok(bytes) = serde_json::to_vec_pretty(&owner) else {
-        return;
+        return false;
     };
-    let _ = fs::write(temp_root.join(OWNER_FILE), bytes);
+    fs::write(temp_root.join(OWNER_FILE), bytes).is_ok()
 }
 
 fn read_browser_process_owner(dir: &Path) -> Option<BrowserProcessOwner> {
@@ -115,34 +121,142 @@ fn trusted_browser_executable(path: &Path) -> bool {
 }
 
 /// 杀前复核：PID 对应的活进程镜像必须就是 owner.json 声明的那个可执行
-/// 文件。仅核对字符串不够——/tmp 全局可写时，同机其他用户可以伪造
-/// owner.json 把 browser_pid 指到本用户的任意进程借清扫杀之。
-/// Linux 进一步要求命令行声明本目录的 profile，把 PID 绑到这份配置
-/// 目录，挡住同用户伪造 owner.json 指向正在使用的浏览器。
+/// 文件，命令行还必须绑定同一份临时 Profile；只核对镜像会把其它 AGC
+/// 实例或用户 Edge 误认成本轮浏览器。
 #[cfg(windows)]
-fn live_process_executable_matches(pid: u32, expected: &Path, _profile_dir: &Path) -> bool {
-    // 已知残余风险：Windows 读他进程命令行成本高（PEB/WMI），此处只核对
-    // 镜像；同用户伪造 owner.json 指向正在使用的浏览器时可误杀它。
+fn windows_process_command_line(pid: u32) -> Option<String> {
+    use std::ffi::c_void;
+    use windows_sys::Win32::Foundation::CloseHandle;
+    use windows_sys::Win32::System::Threading::{
+        OpenProcess, PROCESS_QUERY_INFORMATION, PROCESS_VM_READ,
+    };
+
+    #[repr(C)]
+    struct UnicodeString {
+        length: u16,
+        maximum_length: u16,
+        buffer: *const u16,
+    }
+
+    #[link(name = "ntdll")]
+    unsafe extern "system" {
+        fn NtQueryInformationProcess(
+            process: *mut c_void,
+            information_class: u32,
+            information: *mut c_void,
+            information_length: u32,
+            return_length: *mut u32,
+        ) -> i32;
+    }
+
+    let process = unsafe { OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, 0, pid) };
+    if process.is_null() {
+        return None;
+    }
+    let mut required = 0;
+    let _ =
+        unsafe { NtQueryInformationProcess(process, 60, std::ptr::null_mut(), 0, &mut required) };
+    if required == 0 || required > 64 * 1024 {
+        unsafe { CloseHandle(process) };
+        return None;
+    }
+    let mut buffer = vec![0u8; required as usize];
+    let status = unsafe {
+        NtQueryInformationProcess(
+            process,
+            60,
+            buffer.as_mut_ptr().cast(),
+            required,
+            &mut required,
+        )
+    };
+    unsafe { CloseHandle(process) };
+    if status != 0 {
+        return None;
+    }
+    let command_line = unsafe { &*(buffer.as_ptr().cast::<UnicodeString>()) };
+    if command_line.buffer.is_null() || command_line.length == 0 || command_line.length % 2 != 0 {
+        return None;
+    }
+    let text = unsafe {
+        std::slice::from_raw_parts(command_line.buffer, command_line.length as usize / 2)
+    };
+    String::from_utf16(text).ok()
+}
+
+#[cfg(windows)]
+fn windows_command_line_arguments(command_line: &str) -> Option<Vec<String>> {
+    use windows_sys::Win32::Foundation::LocalFree;
+    use windows_sys::Win32::UI::Shell::CommandLineToArgvW;
+
+    let wide = command_line
+        .encode_utf16()
+        .chain(std::iter::once(0))
+        .collect::<Vec<_>>();
+    let mut count = 0;
+    let argv = unsafe { CommandLineToArgvW(wide.as_ptr(), &mut count) };
+    if argv.is_null() || count < 0 {
+        if !argv.is_null() {
+            unsafe { LocalFree(argv.cast()) };
+        }
+        return None;
+    }
+    let result = unsafe { std::slice::from_raw_parts(argv, count as usize) }
+        .iter()
+        .map(|argument| {
+            if argument.is_null() {
+                return None;
+            }
+            let mut length = 0;
+            unsafe {
+                while *argument.add(length) != 0 {
+                    length += 1;
+                }
+                String::from_utf16(std::slice::from_raw_parts(*argument, length)).ok()
+            }
+        })
+        .collect::<Option<Vec<_>>>();
+    unsafe { LocalFree(argv.cast()) };
+    result
+}
+
+#[cfg(windows)]
+fn command_line_contains_profile(command_line: &str, profile_dir: &Path) -> bool {
+    const MARKER: &str = "--user-data-dir=";
+    let Some(arguments) = windows_command_line_arguments(command_line) else {
+        return false;
+    };
+    arguments.iter().enumerate().any(|(index, argument)| {
+        let value = argument.strip_prefix(MARKER).or_else(|| {
+            (argument == "--user-data-dir")
+                .then(|| arguments.get(index + 1).map(String::as_str))
+                .flatten()
+        });
+        value.is_some_and(|value| same_executable_path(Path::new(value), profile_dir))
+    })
+}
+
+#[cfg(windows)]
+fn live_process_executable_matches(pid: u32, expected: &Path, profile_dir: &Path) -> bool {
     use windows_sys::Win32::Foundation::CloseHandle;
     use windows_sys::Win32::System::Threading::{
         OpenProcess, QueryFullProcessImageNameW, PROCESS_QUERY_LIMITED_INFORMATION,
     };
-    // SAFETY: 句柄非空时由 CloseHandle 释放；打开失败按不匹配处理（fail-closed）。
     let handle = unsafe { OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, 0, pid) };
     if handle.is_null() {
         return false;
     }
     let mut buffer = [0u16; 1024];
     let mut length = buffer.len() as u32;
-    // SAFETY: buffer 可写，length 先传入容量、返回实际长度。
     let okay = unsafe { QueryFullProcessImageNameW(handle, 0, buffer.as_mut_ptr(), &mut length) };
-    // SAFETY: handle 是本函数持有的合法句柄。
     unsafe { CloseHandle(handle) };
     if okay == 0 || length == 0 {
         return false;
     }
     let actual = std::path::PathBuf::from(String::from_utf16_lossy(&buffer[..length as usize]));
     same_executable_path(&actual, expected)
+        && windows_process_command_line(pid)
+            .is_some_and(|command_line| command_line_contains_profile(&command_line, profile_dir))
 }
 
 #[cfg(target_os = "linux")]
@@ -181,7 +295,10 @@ fn directory_owned_by_current_user(dir: &Path) -> bool {
         .unwrap_or(false)
 }
 
-fn kill_browser_process_tree(root_pid: u32, expected_identity: &str) -> Result<(), String> {
+pub(super) fn kill_browser_process_tree(
+    root_pid: u32,
+    expected_identity: &str,
+) -> Result<(), String> {
     #[cfg(windows)]
     {
         // 追踪所有经确认属于这棵树的 PID；只有它们全部从快照中消失才判
@@ -190,15 +307,17 @@ fn kill_browser_process_tree(root_pid: u32, expected_identity: &str) -> Result<(
         for _ in 0..TREE_KILL_MAX_PASSES {
             let snapshot = windows_process_snapshot()?;
             let root_present = snapshot.iter().any(|(pid, _)| *pid == root_pid);
-            if root_present && process_identity_matches(root_pid, expected_identity) {
+            if root_present {
+                if !process_identity_matches(root_pid, expected_identity) {
+                    return Err("browser-sweep-root-identity-mismatch".into());
+                }
                 // root 仍是目标浏览器：发现并追踪当前整棵子树。
                 for pid in windows_process_tree_pids_from(&snapshot, root_pid)? {
                     track_process(&mut tracked, pid);
                 }
-            } else if !root_present {
+            } else {
                 // root 已退出且 PID 未被复用：发现临终前才拉起、仍挂在旧父
-                // PID 上的孤儿子进程。PID 已被复用时不做发现，避免误认
-                // 复用者的子进程。
+                // PID 上的孤儿子进程。PID 已被复用时不会进入此分支。
                 for (pid, ppid) in &snapshot {
                     if *ppid == root_pid {
                         track_process(&mut tracked, *pid);
@@ -323,6 +442,62 @@ fn windows_process_tree_pids_from(
     Ok(tree)
 }
 
+#[cfg(windows)]
+pub(super) fn ensure_browser_tree_in_job(
+    root_pid: u32,
+    root_identity: &str,
+    job: &WindowsProcessJob,
+) -> Result<(), String> {
+    for _ in 0..TREE_KILL_MAX_PASSES {
+        let snapshot = windows_process_snapshot()?;
+        if !snapshot.iter().any(|(pid, _)| *pid == root_pid)
+            || !process_identity_matches(root_pid, root_identity)
+        {
+            return Err("browser-job-root-identity-unconfirmed".into());
+        }
+        let tree = windows_process_tree_pids_from(&snapshot, root_pid)?;
+        if tree.is_empty() {
+            return Err("browser-job-root-exited-before-coverage".into());
+        }
+        let mut members = Vec::with_capacity(tree.len());
+        for pid in tree {
+            let identity = if pid == root_pid {
+                root_identity.to_string()
+            } else {
+                crate::process_identity::external_agent_runner_process_start_identity(pid)
+                    .ok()
+                    .flatten()
+                    .ok_or_else(|| "browser-job-process-identity-unconfirmed".to_string())?
+            };
+            members.push((pid, identity));
+        }
+        for (pid, identity) in members {
+            if !process_identity_matches(pid, &identity) {
+                return Err("browser-job-process-identity-changed".into());
+            }
+            if !job.contains_pid(pid)? {
+                job.assign_pid(pid)?;
+            }
+            if !process_identity_matches(pid, &identity) {
+                return Err("browser-job-process-identity-changed".into());
+            }
+        }
+
+        let verified = windows_process_snapshot()?;
+        let verified_tree = windows_process_tree_pids_from(&verified, root_pid)?;
+        if process_identity_matches(root_pid, root_identity)
+            && !verified_tree.is_empty()
+            && verified_tree
+                .iter()
+                .all(|pid| job.contains_pid(*pid).unwrap_or(false))
+        {
+            return Ok(());
+        }
+        std::thread::sleep(TREE_KILL_PASS_INTERVAL);
+    }
+    Err("browser-job-tree-coverage-unconfirmed".into())
+}
+
 #[cfg(windows)]
 fn windows_terminate_process(pid: u32) {
     use windows_sys::Win32::Foundation::CloseHandle;
@@ -468,6 +643,18 @@ mod tests {
         }
     }
 
+    #[test]
+    fn owner_write_failure_leaves_no_partial_ownership_record() {
+        let root = tempfile::tempdir().unwrap();
+        let missing_root = root.path().join("missing");
+        assert!(!write_browser_process_owner(
+            &missing_root,
+            std::process::id(),
+            &std::env::current_exe().unwrap(),
+        ));
+        assert!(!missing_root.join(OWNER_FILE).exists());
+    }
+
     #[test]
     fn legacy_directory_without_owner_file_is_removed_only_when_old_enough() {
         let root = tempfile::tempdir().unwrap();
@@ -548,6 +735,90 @@ mod tests {
         assert!(trusted_browser_executable(&installed));
     }
 
+    #[cfg(windows)]
+    #[test]
+    fn windows_browser_cleanup_requires_the_exact_profile_argument() {
+        let profile = Path::new(r#"C:\Users\tester\App Data\ga-browser-1\profile"#);
+        assert!(command_line_contains_profile(
+            r#"msedge.exe --headless --user-data-dir="C:\Users\tester\App Data\ga-browser-1\profile""#,
+            profile
+        ));
+        assert!(command_line_contains_profile(
+            r#""msedge.exe" "--user-data-dir=C:\Users\tester\App Data\ga-browser-1\profile""#,
+            profile
+        ));
+        assert!(!command_line_contains_profile(
+            r#"msedge.exe --headless --user-data-dir="C:\Users\tester\App Data\ga-browser-2\profile""#,
+            profile
+        ));
+        assert!(!command_line_contains_profile(
+            r#"msedge.exe --headless --user-data-dir="C:\Users\tester\App Data\ga-browser-1\profile-copy""#,
+            profile
+        ));
+    }
+
+    #[cfg(windows)]
+    #[test]
+    fn browser_job_adopts_children_created_before_assignment() {
+        use std::process::{Command, Stdio};
+
+        let mut child = Command::new("cmd.exe")
+            .args([
+                "/c",
+                "ping",
+                "127.0.0.1",
+                "-n",
+                "60",
+                "&",
+                "ping",
+                "127.0.0.1",
+                "-n",
+                "60",
+            ])
+            .stdin(Stdio::null())
+            .stdout(Stdio::null())
+            .stderr(Stdio::null())
+            .spawn()
+            .expect("spawn fixture");
+        let pid = child.id();
+        let identity = crate::process_identity::external_agent_runner_process_start_identity(pid)
+            .expect("fixture identity")
+            .expect("fixture identity present");
+        std::thread::sleep(Duration::from_millis(500));
+        let job = WindowsProcessJob::assign_std(&child).expect("assign fixture job");
+        ensure_browser_tree_in_job(pid, &identity, &job).expect("adopt fixture tree");
+        let snapshot = windows_process_snapshot().expect("snapshot after adoption");
+        let tree = windows_process_tree_pids_from(&snapshot, pid).expect("tree after adoption");
+        assert!(tree
+            .iter()
+            .all(|member| job.contains_pid(*member).expect("job membership")));
+        job.terminate().expect("terminate fixture job");
+        let _ = child.wait();
+        assert!(job.is_empty().expect("fixture job empty"));
+    }
+
+    #[cfg(windows)]
+    #[test]
+    fn kill_browser_process_tree_rejects_root_identity_mismatch() {
+        use std::process::{Command, Stdio};
+
+        let mut child = Command::new("cmd.exe")
+            .args(["/c", "ping", "127.0.0.1", "-n", "60"])
+            .stdin(Stdio::null())
+            .stdout(Stdio::null())
+            .stderr(Stdio::null())
+            .spawn()
+            .expect("spawn fixture");
+        let pid = child.id();
+        let result = kill_browser_process_tree(pid, "not-the-fixture-identity");
+        let _ = child.kill();
+        let _ = child.wait();
+        assert_eq!(
+            result.expect_err("identity mismatch must fail closed"),
+            "browser-sweep-root-identity-mismatch"
+        );
+    }
+
     #[cfg(windows)]
     #[test]
     fn kill_browser_process_tree_reaps_fixture_tree() {
@@ -600,10 +871,8 @@ mod tests {
     fn live_process_executable_matches_current_process_image() {
         let exe = std::env::current_exe().unwrap();
         let profile = Path::new("C:\\fixture\\ga-browser-x\\profile");
-        // Windows 只核对镜像；Linux 还要求命令行绑定 profile，本测试进程
-        // 不具备该标记，正例只在 Windows 断言。
-        #[cfg(windows)]
-        assert!(live_process_executable_matches(
+        // 当前测试进程不携带目标 Profile 标识，即使镜像一致也不能清理。
+        assert!(!live_process_executable_matches(
             std::process::id(),
             &exe,
             profile

apps/ai-game-creator-shell/src-tauri/src/environment_check/web_creation.rs

@@ -311,6 +311,13 @@ async fn host_npm(
 fn browser_validation_failure_code(error: &str) -> &'static str {
     if error.contains("未发现可用的") {
         "web-preflight-browser-missing"
+    } else if error.starts_with("browser-recovery-") {
+        "web-preflight-browser-recovery-failed"
+    } else if error.contains("browser-temp-unavailable") || error.contains("browser-config-invalid")
+    {
+        "web-preflight-browser-environment-failed"
+    } else if error.starts_with("browser-launch-failed:") {
+        "web-preflight-browser-launch-failed"
     } else if error.contains("启动浏览器超时") {
         "web-preflight-browser-launch-timeout"
     } else if error.contains("启动浏览器失败") {
@@ -342,6 +349,13 @@ fn browser_validation_failure_code(error: &str) -> &'static str {
     }
 }
 
+fn browser_validation_diagnostic(error: &str) -> Option<&str> {
+    (error.starts_with("browser-recovery-")
+        || error.starts_with("browser-launch-failed:")
+        || error.starts_with("browser-cleanup-unconfirmed:"))
+    .then_some(error)
+}
+
 pub(crate) async fn host_web_creation_preflight() -> Value {
     let started = Instant::now();
     // 预检前顺手清扫陈旧的无头浏览器，避免残留进程放大本轮超时。
@@ -362,18 +376,60 @@ pub(crate) async fn host_web_creation_preflight() -> Value {
         let validation = crate::browser::validate_local_preview_in_browser(BrowserValidationInput {
             url: preview.url, viewports: vec![BrowserValidationViewport::Desktop, BrowserValidationViewport::Mobile], expected_text: vec![], settle_ms: 100, fail_on_console_error: true, playtest_scenario: None, evidence_root: root.join("evidence"),
         }).await;
-        let result = validation.map_err(|error| browser_validation_failure_code(&error).to_string())?;
-        if !result.passed || result.viewport_results.len() != 2 { return Err("web-preflight-page-check-failed".into()); }
+        let result = validation
+            .map_err(|error| {
+                let code = browser_validation_failure_code(&error);
+                browser_validation_diagnostic(&error)
+                    .map(|diagnostic| format!("{code};diagnostic={diagnostic}"))
+                    .unwrap_or_else(|| code.to_string())
+            })?;
+        if !result.passed || result.viewport_results.len() != 2 {
+            return Err("web-preflight-page-check-failed".into());
+        }
         let mut pngs = Vec::new();
         for viewport in result.viewport_results {
-            let bytes = fs::read(&viewport.screenshot_path).map_err(|_| "web-preflight-screenshot-missing")?;
-            if !bytes.starts_with(b"\x89PNG\r\n\x1a\n") || bytes.len() < 128 { return Err("web-preflight-screenshot-invalid".into()); }
-            image::load_from_memory(&bytes).map_err(|_| "web-preflight-screenshot-invalid")?;
-            pngs.push(json!({"viewport":viewport.viewport,"passed":viewport.passed,"pngBytes":bytes.len()}));
+            let bytes = fs::read(&viewport.screenshot_path)
+                .map_err(|_| "web-preflight-screenshot-missing")?;
+            if !bytes.starts_with(b"\x89PNG\r\n\x1a\n") || bytes.len() < 128 {
+                return Err("web-preflight-screenshot-invalid".into());
+            }
+            image::load_from_memory(&bytes)
+                .map_err(|_| "web-preflight-screenshot-invalid")?;
+            pngs.push(json!({
+                "viewport": viewport.viewport,
+                "passed": viewport.passed,
+                "pngBytes": bytes.len(),
+            }));
         }
-        Ok::<_, String>(json!({"schemaVersion":"agc-web-creation-preflight.v1","status":"ready","runtime":{"source":runtime.source,"nodeVersion":node,"npmVersion":npm},"build":{"status":"ready","kind":"npm-node-fixture"},"browser":{"status":"ready","viewports":pngs}}))
-    }.await;
-    let mut result = run.unwrap_or_else(|code| json!({"schemaVersion":"agc-web-creation-preflight.v1","status":"blocked","code":code}));
+        Ok::<_, String>(json!({
+            "schemaVersion": "agc-web-creation-preflight.v1",
+            "status": "ready",
+            "runtime": {
+                "source": runtime.source,
+                "nodeVersion": node,
+                "npmVersion": npm,
+            },
+            "build": {"status": "ready", "kind": "npm-node-fixture"},
+            "browser": {"status": "ready", "viewports": pngs},
+        }))
+    }
+    .await;
+    let mut result = run.unwrap_or_else(|error| {
+        let (code, diagnostic) = error
+            .split_once(";diagnostic=")
+            .map_or((error.as_str(), None), |(code, diagnostic)| {
+                (code, Some(diagnostic))
+            });
+        let mut blocked = json!({
+            "schemaVersion": "agc-web-creation-preflight.v1",
+            "status": "blocked",
+            "code": code,
+        });
+        if let Some(diagnostic) = diagnostic {
+            blocked["diagnostic"] = json!(diagnostic);
+        }
+        blocked
+    });
     result["elapsedMs"] = json!(started.elapsed().as_millis());
     result
 }
@@ -382,9 +438,14 @@ pub(crate) async fn host_web_creation_preflight() -> Value {
 pub(crate) async fn preflight_web_game_creation() -> Result<Value, String> {
     let result = host_web_creation_preflight().await;
     if result["status"] != "ready" {
+        let diagnostic = result["diagnostic"]
+            .as_str()
+            .map(|value| format!("；诊断：{value}"))
+            .unwrap_or_default();
         return Err(format!(
-            "Web 游戏环境预检未通过：{}；尚未启动生成",
-            result["code"].as_str().unwrap_or("web-preflight-failed")
+            "Web 游戏环境预检未通过：{}{}；尚未启动生成",
+            result["code"].as_str().unwrap_or("web-preflight-failed"),
+            diagnostic,
         ));
     }
     Ok(result)
@@ -548,16 +609,40 @@ mod tests {
             browser_validation_failure_code("启动浏览器失败：2"),
             "web-preflight-browser-launch-failed"
         );
+
+        assert_eq!(
+            browser_validation_failure_code(
+                "browser-launch-failed: stage=setup cause=browser-temp-unavailable"
+            ),
+            "web-preflight-browser-environment-failed"
+        );
+
+        let recovery_diagnostic = "browser-recovery-failed: initial-stage=ws-handshake final-stage=cdp-connect subprocess-exited=true cleanup-confirmed=true recovery-retried=true";
+        assert_eq!(
+            browser_validation_failure_code(recovery_diagnostic),
+            "web-preflight-browser-recovery-failed"
+        );
+        assert_eq!(
+            browser_validation_diagnostic(recovery_diagnostic),
+            Some(recovery_diagnostic)
+        );
+        assert_eq!(
+            browser_validation_diagnostic("启动浏览器失败：原始错误"),
+            None
+        );
         assert_eq!(
             browser_validation_failure_code("宿主已停止本轮浏览器验证"),
             "web-preflight-cancelled"
         );
+        let cleanup_diagnostic = "browser-cleanup-unconfirmed: stage=shutdown subprocess-exited=true cleanup-confirmed=false recovery-retried=false";
         assert_eq!(
-            browser_validation_failure_code(
-                "browser-cleanup-unconfirmed: 浏览器收束后无法证明退出，请核对本轮验证进程"
-            ),
+            browser_validation_failure_code(cleanup_diagnostic),
             "web-preflight-browser-cleanup-failed"
         );
+        assert_eq!(
+            browser_validation_diagnostic(cleanup_diagnostic),
+            Some(cleanup_diagnostic)
+        );
         assert_eq!(
             browser_validation_failure_code("创建浏览器临时目录失败：拒绝访问"),
             "web-preflight-browser-environment-failed"

apps/ai-game-creator-shell/src-tauri/src/process_session/model.rs

@@ -236,7 +236,42 @@ impl WindowsProcessJob {
             .ok_or_else(|| "子进程缺少 Windows process handle".to_string())?;
         Self::assign_handle(handle as windows_sys::Win32::Foundation::HANDLE)
     }
+    pub(crate) fn contains_pid(&self, pid: u32) -> Result<bool, String> {
+        use windows_sys::Win32::Foundation::CloseHandle;
+        use windows_sys::Win32::System::JobObjects::IsProcessInJob;
+        use windows_sys::Win32::System::Threading::{
+            OpenProcess, PROCESS_QUERY_LIMITED_INFORMATION,
+        };
+        let process = unsafe { OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, 0, pid) };
+        if process.is_null() {
+            return Err("无法打开进程核对 Windows Job 归属".into());
+        }
+        let mut in_job = 0;
+        let okay = unsafe { IsProcessInJob(process, self.0, &mut in_job) };
+        unsafe { CloseHandle(process) };
+        if okay == 0 {
+            return Err("无法核对进程 Windows Job 归属".into());
+        }
+        Ok(in_job != 0)
+    }
 
+    pub(crate) fn assign_pid(&self, pid: u32) -> Result<(), String> {
+        use windows_sys::Win32::Foundation::CloseHandle;
+        use windows_sys::Win32::System::JobObjects::AssignProcessToJobObject;
+        use windows_sys::Win32::System::Threading::{
+            OpenProcess, PROCESS_SET_QUOTA, PROCESS_TERMINATE,
+        };
+        let process = unsafe { OpenProcess(PROCESS_SET_QUOTA | PROCESS_TERMINATE, 0, pid) };
+        if process.is_null() {
+            return Err("无法打开进程加入 Windows Job".into());
+        }
+        let okay = unsafe { AssignProcessToJobObject(self.0, process) };
+        unsafe { CloseHandle(process) };
+        if okay == 0 {
+            return Err("无法将进程加入 Windows Job".into());
+        }
+        Ok(())
+    }
     pub(crate) fn assign_tokio_named(
         child: &tokio::process::Child,
         name: &str,

docs/project-memory/plans/【实施计划】Web预检浏览器失败恢复-2026-10-02.md

@@ -0,0 +1,31 @@
+# Web 预检浏览器失败恢复实现计划
+
+- Version: `1`
+- Status: `active`
+- Date: `2026-10-02`
+- Milestone: [`Web 预检浏览器失败恢复`](./【里程碑】Web预检浏览器失败恢复-2026-10-02.md)
+
+## 实现顺序
+
+1. 将浏览器启动失败建模为带阶段、原因、子进程退出确认、清理确认的内部结果；Windows root spawn 后立即绑定 Job，并把绑定瞬间已经出现的子树逐 PID 纳入同一 Job，失败收束必须复用本轮 `BrowserProcessGuard`，并在成功收束时确认 Windows Job 为空。
+2. 在 Web 预检使用的一次浏览器启动入口加入一次有界恢复：仅 WS/CDP 瞬态失败且首次清理确认后创建新的临时目录/Profile 重试；其余失败直接失败关闭。
+3. 保留/加强 owner 与 sweep 的归属门禁，使用 Windows argv 解析核对完整 Profile 参数；无 owner、身份未知、PID 退出或复用均不得按猜测杀进程。
+4. 将启动失败和清理失败的安全诊断保留到预检 blocked 报告和 Tauri 错误中；稳定错误码独立于诊断文本，前端继续区分宿主阻塞与 IPC 故障。
+5. 补充 Rust 纯策略测试、清理边界测试和现有真实 Edge/Chromium ignored smoke；补充首页错误展示的定向测试。
+
+## 验证命令
+
+- `cargo test --locked --manifest-path apps/ai-game-creator-shell/src-tauri/Cargo.toml --bin genarrative-ai-game-creator-shell browser::`
+- `cargo test --locked --manifest-path apps/ai-game-creator-shell/src-tauri/Cargo.toml --bin genarrative-ai-game-creator-shell environment_check::web_creation::tests`
+- `npx vitest run apps/ai-game-creator-shell/tests/homeWebPreflight.test.tsx`
+- `npm run typecheck --workspace apps/ai-game-creator-shell`
+- `npm run check:encoding`
+- `git diff --check`
+
+可选真实环境证据：安装 Windows Edge 时运行现有 `real_browser_health_checks_can_run_concurrently` 及新增恢复 smoke；无浏览器时保持 ignored，不把缺失环境写成通过。
+
+## 风险与回滚
+
+- Windows 进程命令行读取失败按不匹配处理，不执行杀进程；这可能留下临时目录，但保证不误杀。
+- 首次失败进程树收束未确认时不自动重试，避免第二个 Edge 与残留树并存；错误返回安全诊断。
+- 回滚点为浏览器启动恢复入口和 owner/sweep 归属校验，不触及 Web 预检的 Node/npm 或项目写入流程。

docs/project-memory/plans/【里程碑】Web预检浏览器失败恢复-2026-10-02.md

@@ -0,0 +1,41 @@
+# Web 预检浏览器失败恢复
+
+- Version: `1`
+- Status: `active`
+- Date: `2026-10-02`
+- Parent Spec: [`AI游戏创作智能体 App 实施计划`](../../technical/【技术方案】AI游戏创作智能体App实施计划-2026-06-24.md#自动预检与可信脚手架)
+
+## 目标
+
+修复 Issue #584：Web 游戏环境预检在受控 Edge/Chrome 的 WS 握手或 CDP 连接失败时，能够只针对本轮 AGC 浏览器实例完成清理、使用新的隔离 Profile 重试一次，并在仍失败时保留安全、可行动的诊断信息。
+
+## 边界
+
+- 只改 AGC Rust 浏览器启动/清理、Web 预检错误投影及其定向测试、对应现行专题文档。
+- 清理对象必须同时满足 AGC 临时 Profile、进程启动身份、可信浏览器可执行文件和 Windows 进程树归属；无法确认时 fail-closed。
+- 不执行全量 `taskkill /IM msedge.exe`，不影响用户 Edge、其它 AGC 实例、其它项目或 worktree。
+- 不改变 Web 预检的失败关闭、Node/npm 检查、桌面/移动双视口检查和首次生成顺序。
+- 不新增网络、自动下载、跳过浏览器验证或伪造 ready 的旁路。
+
+## 行为合同
+
+1. WS 握手失败、WS 超时、CDP 连接失败或 CDP 连接超时属于可恢复的浏览器启动瞬态失败。
+2. 第一次失败后，宿主必须先确认本轮进程树已退出并清理本轮 Profile / owner 记录 / 临时目录；清理未确认时不得启动第二次浏览器。
+3. 清理确认后，第二次启动必须创建新的隔离临时目录和 Profile；第二次成功后继续现有 desktop/mobile 预检。
+4. 连续失败或清理被阻断时，结果保持 blocked，并带有阶段（WS 握手或 CDP 连接）、子进程是否退出、清理是否确认、是否执行恢复重试等安全诊断。
+5. owner.json 缺失/写入失败、目录过新、进程已退出、PID 被复用、身份未知或归属不明时只能按保守路径处理：不杀不明进程；可安全删除的临时目录才删除。
+
+## 验收标准
+
+- 定向测试构造 `browser-ws-failed` 后证明只在清理确认时重试，且重试使用新的 Profile；第二次成功返回成功。
+- 定向测试覆盖 WS/CDP 阶段、连续失败、清理未确认、owner.json 缺失/无效、刚创建目录、进程退出、PID 复用和非 AGC 进程跳过。
+- Windows 真实 Edge 安装版 smoke 保留在现有真实浏览器测试入口中；无 Edge 的环境明确跳过而不是伪造通过。
+- 现有首页预检、正常 Edge 使用、首次生成失败关闭和前端 IPC/宿主错误区分回归通过。
+
+## 依赖
+
+- `apps/ai-game-creator-shell/src-tauri/src/browser/process.rs`
+- `apps/ai-game-creator-shell/src-tauri/src/browser/sweep.rs`
+- `apps/ai-game-creator-shell/src-tauri/src/environment_check/web_creation.rs`
+- `apps/ai-game-creator-shell/src/features/app-shell/homeWebPreflight.ts`
+- 对应 Rust 与 Vitest 测试。

docs/technical/【技术方案】AI游戏创作智能体App实施计划-2026-06-24.md

@@ -134,7 +134,7 @@ UI 编辑器的“分析参考图”步骤、Rust 命令 `suggest_ui_design_sema
 - 对客户端刚创建且内容仍匹配可信模板的 Web 脚手架，在正式生成前由宿主执行受控依赖准备和真实 Vite 构建。依赖安装禁用生命周期脚本；不自动安装或覆盖导入/用户修改过的工程。
 - 准备凭证的 `ready` 只证明初次环境准备成功，不代表当前游戏已验收，后续正常修改不得因此重新安装。`preparing` 中断恢复必须证明原拥有者已结束且其执行子树已回收；身份或归属未知时保留阻断，不重复执行。
 - 输出明确区分“Node/npm 构建能力通过”与“本项目 Vite 构建通过”；任何一步失败保留可行动错误，不降级为预检成功。
-- 首页必须区分宿主预检阻塞与 Tauri IPC 调用失败：Rust 返回的安全错误码可透传到状态栏；无错误码的瞬态 IPC 失败只允许一次有界重试后显示独立的客户端连接故障，不得统一伪装成 Node/npm 或浏览器故障。预检仍保持失败关闭，不得因为展示错误变得可绕过。
+- 首页必须区分宿主预检阻塞与 Tauri IPC 调用失败：Rust 返回的安全错误码可透传到状态栏；无错误码的瞬态 IPC 失败只允许一次有界重试后显示独立的客户端连接故障，不得统一伪装成 Node/npm 或浏览器故障。预检仍保持失败关闭，不得因为展示错误变得可绕过。浏览器 WS 握手、WS 超时和 CDP 连接失败只允许在确认本轮 AGC 浏览器进程树已退出、Profile/owner/临时目录已按归属清理后使用新的隔离 Profile 自动恢复一次；清理未确认时不得重试，连续失败必须返回包含阶段、子进程退出确认、清理确认和是否重试的安全诊断。
 - 安装载荷提供相同的安全预检 CLI，验证无系统 Node 的独立运行、缺包/篡改失败关闭。NSIS 解包载荷 smoke 与真实安装器注册流程分开报告，不覆盖当前用户的既有安装。
 
 ### 跑酷固定基线
@@ -188,7 +188,7 @@ UI 编辑器的“分析参考图”步骤、Rust 命令 `suggest_ui_design_sema
 ### 环境与工作流
 
 - 客户端交付配套 Node/npm；发布包从本机已安装且与目标平台/架构一致的工具链制作受校验资源，保留许可并校验内容摘要。安装态不依赖系统 PATH 的 Node；开发态可使用已验证的宿主运行时。不得从项目或相对 PATH 加载伪造运行时。随包运行时是**单架构**官方发行版，因此 macOS 当前只构建 `aarch64-apple-darwin` 单架构包；要出 universal 必须先让 staging 支持按架构各带一份同版本运行时，在此之前 universal 目标失败关闭，不得只带宿主架构那一份糊过去。
-- 新建 Web 游戏在生图和大量实现前执行客户端环境预检，检查 Node/npm 的实际版本、浏览器启动和 CDP 可用性。报告只包含安全状态、版本、耗时和错误码；错误码必须按真实原因分流，浏览器验证只允许在确有证据时使用 `web-preflight-browser-missing` / `-launch-timeout` / `-launch-failed` / `-browser-cleanup-failed`，取消、证据写入、输入与页面校验各有独立码，未识别原因落回专用 `web-preflight-unclassified`，不得用一个具体子系统码兜底。缺失或异常必须尽早返回阻塞，不能指示模型改宿主环境、全盘搜索或自行下载一套运行时。编辑器工程不强制 Web 工具链。
+- 新建 Web 游戏在生图和大量实现前执行客户端环境预检，检查 Node/npm 的实际版本、浏览器启动和 CDP 可用性。报告只包含安全状态、版本、耗时和错误码；错误码必须按真实原因分流，浏览器验证只允许在确有证据时使用 `web-preflight-browser-missing` / `-launch-timeout` / `-launch-failed` / `-browser-environment-failed` / `-browser-cleanup-failed` / `-browser-recovery-failed`，取消、证据写入、输入与页面校验各有独立码，未识别原因落回专用 `web-preflight-unclassified`，不得用一个具体子系统码兜底。浏览器恢复诊断只保留阶段、进程退出、清理确认和重试状态等安全字段，不暴露命令行、路径或上游原文。缺失或异常必须尽早返回阻塞，不能指示模型改宿主环境、全盘搜索或自行下载一套运行时。编辑器工程不强制 Web 工具链。
 - 预检不安装依赖、不修改项目 revision、不请求平台生成；构建仍执行项目自己的 npm 脚本。Codex 隔离 HOME 与平台凭据边界保持不变，客户端把已验证的运行时加入执行 PATH，不能把宿主凭据目录交给模型。
 - 第一轮先明确本次必需玩法、素材和验收项。同批独立读取尽量合并，必需图片一次规划；已有且可用的资产复用。已有目标全部通过后给出交付结果，非阻塞的新点子列为后续工作，不在收尾时主动开启新的生产链。