fix: harden trans playwright passthrough

AGENTS.md

@@ -255,7 +255,7 @@ UniDesk 是一个以主 server 为统一入口的分布式工作平台；本文
 - `bun scripts/cli.ts codex judge <taskId> --attempt <n> [--dry-run]`：按指定 task/attempt 用与队列 worker 相同的上下文构建和 MiniMax judge 调用路径单步复现完成判定；`--dry-run` 只输出 prompt/payload 诊断。
 - `bun scripts/cli.ts codex steer <taskId> [prompt|--prompt-file path|--prompt-stdin] [--steer-id id] [--dry-run] [--no-retry|--retry-attempts N]` / `codex steer-confirm <taskId> --steer-id <id>`：向运行中的 active turn 注入纠偏提示并用 `steerId` 做幂等/trace 确认；真实输出不回显 prompt，遇到 `deliveryUnconfirmed` 先查确认命令，不重复发送同一纠偏。
 - `bun scripts/cli.ts codex interrupt|cancel <taskId>`：通过 Code Queue 私有代理中断运行任务或取消 queued/retry_wait 任务，规则见 `docs/reference/cli.md`。
-- `bun scripts/playwright-cli.ts screenshot|open|eval ...` / `trans D601 playwright <<'PW' ...`：UniDesk 仓库自带的 Playwright 指挥手测 wrapper 与跨 host 透传入口，默认 headless，支持 heredoc 执行并把截图等产物回传到本机 `/tmp`；不实现长驻浏览器 daemon，规则见 `docs/reference/cli.md`。
+- `bun scripts/playwright-cli.ts screenshot|open|eval ...` / `trans D601 playwright <<'PW' ...`：UniDesk 仓库自带的 Playwright 指挥手测 wrapper 与跨 host 透传入口，默认 headless，支持 heredoc 后台执行、短轮询和截图等产物回传到本机 `/tmp`；不实现长驻浏览器 daemon，规则见 `docs/reference/cli.md`。
 - `bun scripts/cli.ts server stop`：以异步 job 停止固定 Compose 项目中的全部 UniDesk 服务，停止后用 `server status` 复核。
 - `bun scripts/cli.ts job list [--limit N]` / `bun scripts/cli.ts job status latest [--tail-bytes N]`：分页查询 `.state/jobs/` 中的异步任务状态，状态输出只读日志尾部并保留完整日志路径，job 机制见 `docs/reference/cli.md`。
 - `bun scripts/cli.ts debug health` / `bun scripts/cli.ts debug dispatch` / `bun scripts/cli.ts debug task`：通过 Docker 内网 core、真实 HTTP、WebSocket、系统指标、Docker 状态和 Host SSH 维护桥流程调试健康检查、任务下发与任务结果，调试规则见 `docs/reference/cli.md`。

docs/reference/cli.md

@@ -146,9 +146,9 @@ playwright-cli screenshot https://example.com "$UNIDESK_PLAYWRIGHT_SCREENSHOT" -
 PW
 ```
 
-`playwright` operation 读取 stdin heredoc，在目标 POSIX host/workload 上临时注入 `playwright-cli` wrapper 到 `PATH`。wrapper 优先使用远端用户的 `~/.agents/skills/playwright*/scripts/playwright-cli.ts`，其次使用 route workspace 中的 `./scripts/playwright-cli.ts`，最后才使用远端 `PATH` 中的 `playwright-cli`。命令会设置 `UNIDESK_PLAYWRIGHT_REMOTE_DIR` 和 `UNIDESK_PLAYWRIGHT_SCREENSHOT`，把远端 run 目录中的 `png/jpg/jpeg/webp/pdf` 产物回传到本机 `--local-dir`，默认 `/tmp`，并返回本地路径、远端路径、字节数、SHA-256、stdout/stderr tail 和 transfer verification。
+`playwright` operation 读取 stdin heredoc，在目标 POSIX host/workload 上临时注入 `playwright-cli` wrapper 到 `PATH`。wrapper 优先使用 route workspace 或目标 host 上已知 UniDesk workspace 的 `./scripts/playwright-cli.ts`，其次使用远端用户的 `~/.agents/skills/playwright*/scripts/playwright-cli.ts`，最后才使用远端 `PATH` 中的 `playwright-cli`。命令会设置 `UNIDESK_PLAYWRIGHT_REMOTE_DIR` 和 `UNIDESK_PLAYWRIGHT_SCREENSHOT`，把远端 run 目录中的 `png/jpg/jpeg/webp/pdf` 产物回传到本机 `--local-dir`，默认 `/tmp`，并返回本地路径、远端路径、字节数、SHA-256、stdout/stderr tail 和 transfer verification。
 
-该入口只负责短生命周期 Playwright 执行和产物回传，不提供长驻浏览器 daemon。需要多步交互时，把步骤写在同一个 heredoc 内；需要保留远端证据时显式加 `--keep-remote`。
+该入口只负责短生命周期 Playwright 执行和产物回传，不提供长驻浏览器 daemon。需要多步交互时，把步骤写在同一个 heredoc 内；helper 会在远端后台提交 job，并用短连接轮询 manifest，避免单次 SSH 透传超过 60 秒硬限制。需要保留远端证据时显式加 `--keep-remote`。
 
 ## Async Job State
 

scripts/src/help.ts

@@ -210,7 +210,7 @@ export function sshHelp(): unknown {
       "For arbitrary stdin streams into a workload command, use a workload route plus `exec --stdin -- <command> ...`; this keeps the route as location-only and avoids heredoc/base64/tar shell wrapping.",
       "`apply-patch` is the default remote text patch entry and uses the v2 local line-based patch engine with remote read/write operations, including Windows routes such as `D601:win/c/test`, so long Unicode/Chinese lines and pure insertion hunks avoid the legacy remote shell hunk parser. Plain multi-file Update File patches on POSIX host/k3s and Windows workspace routes use bulk read/write operations to avoid per-file SSH round trips. Its stdout follows Codex apply_patch text output rather than UniDesk JSON output; stderr keeps Codex-style failure text and appends one `UNIDESK_APPLY_PATCH_TIMING` JSON summary with durationMs, patchBytes, fileCount, hunkCount, changedCount, remoteOperationCount, remoteOperationCounts and remoteElapsedMs so slow patch runs can be attributed without changing success stdout.",
       "`upload` and `download` are the default whole-file transfer entries for non-text and generated files. They write through remote temp files, verify byte count and SHA-256 on both sides, and return `verification.automatic=true`, `verification.verified=true`, and `verification.match.{bytes,sha256}=true`; this JSON is the transfer integrity proof, so callers do not need a separate manual `sha256sum` check. Downloads stream over `host.ssh.tcp-pool`, emit progress JSON, and may receive a caller-supplied `--inactivity-timeout-ms` from async artifact/deploy jobs so active large transfers are not killed by the generic short-command budget.",
-      "`playwright` runs a stdin heredoc on the target POSIX host/workload with a temporary `playwright-cli` wrapper in PATH, sets `UNIDESK_PLAYWRIGHT_SCREENSHOT` under a remote `/tmp` run directory, then downloads image/pdf artifacts to local `/tmp` by default with the same verified SHA-256 transfer path. Canonical syntax is `trans D601 playwright <<'PW' ... PW`; workspace routes are optional and only affect wrapper discovery.",
+      "`playwright` runs a stdin heredoc on the target POSIX host/workload with a temporary `playwright-cli` wrapper in PATH, submits the remote script as a background job, polls short status commands for the manifest, then downloads image/pdf artifacts to local `/tmp` by default with the same verified SHA-256 transfer path. Canonical syntax is `trans D601 playwright <<'PW' ... PW`; workspace and known UniDesk host workspaces are preferred for wrapper discovery before external skill passthroughs.",
       "`apply-patch-v1` is the only legacy fallback entry: it rejects low-context update hunks by default, reports the matched file:line for each hunk on stderr, and only accepts --allow-loose when the caller has manually reviewed an intentionally ambiguous insertion.",
       "script defaults to target /bin/sh and inherits provider proxy variables such as HTTP_PROXY/HTTPS_PROXY/ALL_PROXY/NO_PROXY; it is for host/k3s POSIX shell only. Use --shell bash only for bash syntax such as pipefail, arrays, or [[ ... ]], not as a proxy workaround.",
       "Route syntax is `{provider}:{plane}[:{scope...}] {operation} [operation-args...]`: the first argv token locates a distributed target only, and every following token belongs to the operation parser. Host workspace routes use `<provider>:/absolute/workspace`; WSL providers can use `<provider>:win ps` for Windows PowerShell and `<provider>:win cmd` for Windows cmd.exe, with `<provider>:win/c/test ...` mapping the Windows cwd to `C:\\test`; native k3s providers such as D601 and G14 use `<provider>:k3s` for the control plane and `<provider>:k3s:<namespace>:<workload>[:<container>]` for a workload/container. In k3s routes, `:` is the distributed route separator; `/...` is only an in-container filesystem cwd and never selects a container. Prefer operation `--cwd /path` when a container is also specified.",

scripts/src/ssh-playwright.ts

@@ -13,6 +13,8 @@ interface SshPlaywrightOptions {
   remoteDir: string | null;
   keepRemote: boolean;
   inactivityTimeoutMs?: number;
+  pollIntervalMs: number;
+  waitTimeoutMs: number;
 }
 
 interface RemotePlaywrightManifest {
@@ -62,9 +64,12 @@ export async function runSshPlaywrightOperation(
   const runId = `unidesk-playwright-${safePathSegment(invocation.providerId)}-${Date.now()}-${randomBytes(4).toString("hex")}`;
   const remoteDir = options.remoteDir ?? `/tmp/${runId}`;
   const localDir = resolve(options.localDir);
-  const remoteCommand = builders.buildRouteCommand(invocation.route, ["sh", "-c", remotePlaywrightRunnerScript(remoteDir), "unidesk-playwright"], { stdin: true });
-  const result = await executor.runRemoteCommand(remoteCommand, userScript);
-  const manifest = parseRemoteManifest(result, remoteDir, runId);
+  const submitCommand = builders.buildRouteCommand(invocation.route, ["sh", "-c", remotePlaywrightSubmitScript(remoteDir), "unidesk-playwright-submit"], { stdin: true });
+  const submit = await executor.runRemoteCommand(submitCommand, userScript);
+  if (submit.exitCode !== 0) {
+    throw new Error(`ssh playwright submit failed: exitCode=${submit.exitCode}; stdoutTail=${JSON.stringify(submit.stdout.slice(-1000))}; stderrTail=${JSON.stringify(submit.stderr.slice(-1000))}`);
+  }
+  const manifest = await pollRemoteManifest(invocation.route, executor, builders, remoteDir, runId, options);
   const artifacts: Array<SshVerifiedDownloadResult & { manifestBytes: number | null; manifestSha256: string | null }> = [];
   let downloadFailure: Record<string, unknown> | null = null;
 
@@ -117,9 +122,10 @@ export async function runSshPlaywrightOperation(
     cleanup,
     downloadFailure,
     remoteCommand: {
-      exitCode: result.exitCode,
-      stdoutBytes: Buffer.byteLength(result.stdout, "utf8"),
-      stderrBytes: Buffer.byteLength(result.stderr, "utf8"),
+      exitCode: manifest.exitCode,
+      submitExitCode: submit.exitCode,
+      submitStdoutBytes: Buffer.byteLength(submit.stdout, "utf8"),
+      submitStderrBytes: Buffer.byteLength(submit.stderr, "utf8"),
     },
   };
   process.stdout.write(`${JSON.stringify(payload, null, 2)}\n`);
@@ -132,6 +138,8 @@ function parseSshPlaywrightOptions(args: string[]): SshPlaywrightOptions {
     localDir: "/tmp",
     remoteDir: null,
     keepRemote: false,
+    pollIntervalMs: 2_000,
+    waitTimeoutMs: 180_000,
   };
   for (let index = 1; index < args.length; index += 1) {
     const arg = args[index] ?? "";
@@ -171,6 +179,24 @@ function parseSshPlaywrightOptions(args: string[]): SshPlaywrightOptions {
       options.inactivityTimeoutMs = parsePositiveInteger("--inactivity-timeout-ms", arg.slice("--inactivity-timeout-ms=".length));
       continue;
     }
+    if (arg === "--poll-interval-ms") {
+      options.pollIntervalMs = parsePositiveInteger(arg, next);
+      index += 1;
+      continue;
+    }
+    if (arg.startsWith("--poll-interval-ms=")) {
+      options.pollIntervalMs = parsePositiveInteger("--poll-interval-ms", arg.slice("--poll-interval-ms=".length));
+      continue;
+    }
+    if (arg === "--wait-timeout-ms") {
+      options.waitTimeoutMs = parsePositiveInteger(arg, next);
+      index += 1;
+      continue;
+    }
+    if (arg.startsWith("--wait-timeout-ms=")) {
+      options.waitTimeoutMs = parsePositiveInteger("--wait-timeout-ms", arg.slice("--wait-timeout-ms=".length));
+      continue;
+    }
     throw new Error(`unsupported ssh playwright option: ${arg}`);
   }
   return options;
@@ -190,7 +216,8 @@ function playwrightHelp(): Record<string, unknown> {
     ],
     behavior: [
       "Reads a POSIX shell heredoc from stdin and runs it on the target route.",
-      "Prepends a temporary playwright-cli wrapper to PATH. The wrapper prefers ~/.agents/skills/playwright*/scripts/playwright-cli.ts, then ./scripts/playwright-cli.ts in the route workspace, then a playwright-cli binary.",
+      "Prepends a temporary playwright-cli wrapper to PATH. The wrapper prefers ./scripts/playwright-cli.ts in the route workspace or known UniDesk host workspaces, then ~/.agents/skills/playwright*/scripts/playwright-cli.ts, then a playwright-cli binary.",
+      "Submits the remote script as a background job and polls short status commands for the manifest, so multi-step Playwright flows do not occupy one SSH connection past the 60s trans runtime limit.",
       "Sets UNIDESK_PLAYWRIGHT_REMOTE_DIR and UNIDESK_PLAYWRIGHT_SCREENSHOT. Files created under that remote dir with image/pdf extensions are downloaded to --local-dir with byte and sha256 verification.",
     ],
     options: {
@@ -198,6 +225,8 @@ function playwrightHelp(): Record<string, unknown> {
       "--remote-dir <absolute-path>": "Remote artifact directory. Default: /tmp/unidesk-playwright-<provider>-<timestamp>-<id>.",
       "--keep-remote": "Do not remove the remote artifact directory after the transfer.",
       "--inactivity-timeout-ms <n>": "Forwarded to verified artifact download when large screenshots are returned.",
+      "--wait-timeout-ms <n>": "Maximum wall-clock time to wait for the remote Playwright job. Default: 180000.",
+      "--poll-interval-ms <n>": "Short-query polling interval for the remote job. Default: 2000.",
     },
   };
 }
@@ -227,17 +256,94 @@ async function readAllStdin(): Promise<string> {
   return Buffer.concat(chunks).toString("utf8");
 }
 
+async function pollRemoteManifest(
+  route: ParsedSshRoute,
+  executor: SshRemoteCommandExecutor,
+  builders: SshFileTransferCommandBuilders,
+  remoteDir: string,
+  runId: string,
+  options: SshPlaywrightOptions,
+): Promise<RemotePlaywrightManifest> {
+  const deadline = Date.now() + options.waitTimeoutMs;
+  let lastStatus: SshCaptureResult | null = null;
+  while (Date.now() <= deadline) {
+    const command = builders.buildRouteCommand(route, ["sh", "-c", remotePlaywrightStatusScript(), "unidesk-playwright-status", remoteDir]);
+    const status = await executor.runRemoteCommand(command);
+    lastStatus = status;
+    if (status.exitCode === 0 && status.stdout.includes(manifestEnd)) return parseRemoteManifest(status, remoteDir, runId);
+    if (status.exitCode !== 0 && !/status\t(?:pending|running)/u.test(status.stdout)) {
+      throw new Error(`ssh playwright status failed: exitCode=${status.exitCode}; stdoutTail=${JSON.stringify(status.stdout.slice(-1000))}; stderrTail=${JSON.stringify(status.stderr.slice(-1000))}`);
+    }
+    await sleep(options.pollIntervalMs);
+  }
+  throw new Error(`ssh playwright timed out waiting for remote job after ${options.waitTimeoutMs}ms; lastStdoutTail=${JSON.stringify(lastStatus?.stdout.slice(-1000) ?? "")}; lastStderrTail=${JSON.stringify(lastStatus?.stderr.slice(-1000) ?? "")}`);
+}
+
+function sleep(ms: number): Promise<void> {
+  return new Promise((resolveSleep) => setTimeout(resolveSleep, ms));
+}
+
+function remotePlaywrightSubmitScript(remoteDir: string): string {
+  const runner = Buffer.from(remotePlaywrightRunnerScript(remoteDir), "utf8").toString("base64");
+  return [
+    "set -eu",
+    `remote_dir=${shellQuote(remoteDir)}`,
+    'mkdir -p -- "$remote_dir"',
+    'user_script="$remote_dir/user-script.sh"',
+    'runner_script="$remote_dir/runner.sh"',
+    'pid_file="$remote_dir/pid"',
+    'submit_stdout="$remote_dir/submit.out"',
+    'submit_stderr="$remote_dir/submit.err"',
+    'if [ -f "$pid_file" ] && kill -0 "$(cat "$pid_file")" >/dev/null 2>&1; then',
+    '  printf "status\\trunning\\nremote_dir\\t%s\\npid\\t%s\\n" "$remote_dir" "$(cat "$pid_file")"',
+    "  exit 0",
+    "fi",
+    'cat > "$user_script"',
+    'chmod 700 "$user_script"',
+    `printf %s ${shellQuote(runner)} | base64 -d >"$runner_script"`,
+    'chmod 700 "$runner_script"',
+    'nohup sh "$runner_script" "$user_script" >"$submit_stdout" 2>"$submit_stderr" < /dev/null &',
+    'pid=$!',
+    'printf "%s\\n" "$pid" >"$pid_file"',
+    'printf "status\\tsubmitted\\nremote_dir\\t%s\\npid\\t%s\\n" "$remote_dir" "$pid"',
+  ].join("\n");
+}
+
+function remotePlaywrightStatusScript(): string {
+  return [
+    "set -eu",
+    'remote_dir="$1"',
+    'manifest="$remote_dir/manifest.tsv"',
+    'pid_file="$remote_dir/pid"',
+    'if [ -f "$manifest" ]; then cat "$manifest"; exit 0; fi',
+    'if [ -f "$pid_file" ] && kill -0 "$(cat "$pid_file")" >/dev/null 2>&1; then',
+    '  printf "status\\trunning\\nremote_dir\\t%s\\npid\\t%s\\n" "$remote_dir" "$(cat "$pid_file")"',
+    "  exit 1",
+    "fi",
+    'if [ -f "$pid_file" ]; then',
+    '  printf "status\\texited-without-manifest\\nremote_dir\\t%s\\npid\\t%s\\n" "$remote_dir" "$(cat "$pid_file")"',
+    '  if [ -f "$remote_dir/submit.err" ]; then tail -c 1000 "$remote_dir/submit.err" >&2; fi',
+    '  if [ -f "$remote_dir/stderr.log" ]; then tail -c 1000 "$remote_dir/stderr.log" >&2; fi',
+    "  exit 2",
+    "fi",
+    'printf "status\\tpending\\nremote_dir\\t%s\\n" "$remote_dir"',
+    'if [ -f "$remote_dir/submit.err" ]; then tail -c 1000 "$remote_dir/submit.err" >&2; fi',
+    'if [ -f "$remote_dir/stderr.log" ]; then tail -c 1000 "$remote_dir/stderr.log" >&2; fi',
+    "exit 1",
+  ].join("\n");
+}
+
 function remotePlaywrightRunnerScript(remoteDir: string): string {
   return [
     "set -eu",
     `remote_dir=${shellQuote(remoteDir)}`,
+    'if [ "$#" -lt 1 ]; then printf "missing user script path\\n" >&2; exit 2; fi',
+    'user_script="$1"',
     'mkdir -p -- "$remote_dir/bin"',
-    'user_script="$remote_dir/user-script.sh"',
     'stdout_file="$remote_dir/stdout.log"',
     'stderr_file="$remote_dir/stderr.log"',
     'artifacts_file="$remote_dir/artifacts.tsv"',
-    'cat > "$user_script"',
-    'chmod 700 "$user_script"',
+    'manifest_file="$remote_dir/manifest.tsv"',
     "sha256_file() {",
     "  if command -v sha256sum >/dev/null 2>&1; then sha256sum -- \"$1\" | awk '{print $1}'; return; fi",
     "  if command -v shasum >/dev/null 2>&1; then shasum -a 256 -- \"$1\" | awk '{print $1}'; return; fi",
@@ -245,14 +351,21 @@ function remotePlaywrightRunnerScript(remoteDir: string): string {
     "  printf 'missing sha256 tool\\n' >&2; return 127",
     "}",
     "resolve_playwright_cli() {",
+    "  if command -v bun >/dev/null 2>&1; then",
+    "    if [ -f ./scripts/playwright-cli.ts ]; then",
+    "      UNIDESK_PLAYWRIGHT_CLI_MODE=repo; UNIDESK_PLAYWRIGHT_CLI_CWD=$(pwd); UNIDESK_PLAYWRIGHT_CLI_BIN=; return 0",
+    "    fi",
+    "    for dir in \"$HOME/workspace/unidesk-dev\" \"$HOME/unidesk\" \"/home/ubuntu/workspace/unidesk-dev\" \"/root/unidesk\"; do",
+    "      if [ -f \"$dir/scripts/playwright-cli.ts\" ]; then",
+    "        UNIDESK_PLAYWRIGHT_CLI_MODE=repo; UNIDESK_PLAYWRIGHT_CLI_CWD=$dir; UNIDESK_PLAYWRIGHT_CLI_BIN=; return 0",
+    "      fi",
+    "    done",
+    "  fi",
     "  for dir in \"$HOME/.agents/skills/playwright\" \"$HOME/.agents/skills/playwright-cli\" \"$HOME/.codex/skills/playwright\" \"$HOME/.codex/skills/playwright-cli\"; do",
     "    if command -v bun >/dev/null 2>&1 && [ -f \"$dir/scripts/playwright-cli.ts\" ]; then",
     "      UNIDESK_PLAYWRIGHT_CLI_MODE=skill; UNIDESK_PLAYWRIGHT_CLI_CWD=$dir; UNIDESK_PLAYWRIGHT_CLI_BIN=; return 0",
     "    fi",
     "  done",
-    "  if command -v bun >/dev/null 2>&1 && [ -f ./scripts/playwright-cli.ts ]; then",
-    "    UNIDESK_PLAYWRIGHT_CLI_MODE=repo; UNIDESK_PLAYWRIGHT_CLI_CWD=$(pwd); UNIDESK_PLAYWRIGHT_CLI_BIN=; return 0",
-    "  fi",
     "  if command -v playwright-cli >/dev/null 2>&1; then",
     "    UNIDESK_PLAYWRIGHT_CLI_MODE=bin; UNIDESK_PLAYWRIGHT_CLI_CWD=; UNIDESK_PLAYWRIGHT_CLI_BIN=$(command -v playwright-cli); return 0",
     "  fi",
@@ -286,7 +399,6 @@ function remotePlaywrightRunnerScript(remoteDir: string): string {
     "set +e",
     'sh "$user_script" >"$stdout_file" 2>"$stderr_file"',
     "user_rc=$?",
-    "set -e",
     ': > "$artifacts_file"',
     'find "$remote_dir" -type f | while IFS= read -r file; do',
     '  case "$file" in "$user_script"|"$stdout_file"|"$stderr_file"|"$artifacts_file"|"$remote_dir/bin/"*) continue ;; esac',
@@ -298,6 +410,7 @@ function remotePlaywrightRunnerScript(remoteDir: string): string {
     "  esac",
     "done",
     "b64_tail() { if [ -f \"$1\" ]; then tail -c 4000 \"$1\" | base64 | tr -d '\\n'; fi; }",
+    `{`,
     `printf '%s\\n' ${shellQuote(manifestBegin)}`,
     "printf 'run_id\\t%s\\n' \"${remote_dir##*/}\"",
     "printf 'exit_code\\t%s\\n' \"$user_rc\"",
@@ -312,6 +425,7 @@ function remotePlaywrightRunnerScript(remoteDir: string): string {
     "printf 'stderr_tail_b64\\t%s\\n' \"$(b64_tail \"$stderr_file\")\"",
     'cat "$artifacts_file"',
     `printf '%s\\n' ${shellQuote(manifestEnd)}`,
+    `} > "$manifest_file"`,
     'exit "$user_rc"',
   ].join("\n");
 }