fix(tools): Bash 结果显式说明被 provider 超时上限杀掉

crates/agent-gui/src/lib/tools/shellTools.ts

@@ -734,10 +734,32 @@ export function createShellTools(params: {
     stdout: string;
     stderr: string;
     shellFamily?: string;
+    /** 本次运行是否被 shell 超时杀掉（`ShellRunResponse.timed_out`）。 */
+    timedOut?: boolean;
+    /** 实际生效的超时上限（res.effective_timeout_ms，回退到请求值）。 */
+    effectiveTimeoutMs?: number;
+    /** 当前 provider 的硬上限；等于 effectiveTimeoutMs 时说明是策略上限而非请求值。 */
+    timeoutCapMs?: number;
+    providerLabel?: string;
   }) {
     const combined = [params.command, params.stdout, params.stderr].join("\n");
     const hints: string[] = [];
 
+    if (params.timedOut) {
+      // 报告里的真实困惑："同一个构建为什么连续跑了三次、日志里看不到失败原因"。
+      // 超时杀进程时，命令自己的日志文件里当然什么都没有——它被外部 SIGKILL 了。
+      // 所以这里把"被杀的原因"写成模型和人都能直接读到的一行，并说明重跑同一
+      // 条命令不会有别的结果（这一层从来没有自动重试，重复执行都是模型自己发的）。
+      const limitMs = params.effectiveTimeoutMs ?? 0;
+      const capMs = params.timeoutCapMs ?? limitMs;
+      hints.push(
+        `Hint: This run was killed by the shell timeout (${limitMs}ms) — it did not crash, and the command itself reported no error. ` +
+          `The partial output above is everything it produced before the kill. ` +
+          `${params.providerLabel ?? "This provider"} caps Bash at ${capMs}ms, so re-running the identical command unchanged reaches the same limit: ` +
+          `redirect the output to a file and poll that file, split the work into steps that finish within ${capMs}ms, or pass an explicit timeout_ms (clamped to ${capMs}ms).`,
+      );
+    }
+
     if (
       runtimePlatform === "windows" &&
       (params.shellFamily === "powershell" || params.shellFamily === "cmd")
@@ -1582,6 +1604,10 @@ export function createShellTools(params: {
               stdout: res.stdout || "",
               stderr: res.stderr || "",
               shellFamily: res.shell_family,
+              timedOut: Boolean(res.timed_out),
+              effectiveTimeoutMs: res.effective_timeout_ms || timeout_ms,
+              timeoutCapMs: timeoutPolicy.maxTimeoutMs,
+              providerLabel: timeoutPolicy.providerLabel,
             })
           : "";
 

crates/agent-gui/test/tools/shell-tools.test.mjs

@@ -1520,3 +1520,54 @@ test("resumable Bash blocks leading sleep polling but allows short or internal s
   );
   assert.equal(calls.length, 2);
 });
+
+test("Bash tool spells out a provider-capped timeout instead of leaving the kill unexplained", async () => {
+  // 实测报告里的困惑："同一个构建连续跑了三次，前两次日志里看不到失败原因"。
+  // 这一层从来没有自动重试——重复执行都是模型自己发起的；前两次是被 provider
+  // 的 30s 上限杀掉的，而被外部杀死的命令自己的日志里当然什么都没有。所以工具
+  // 结果必须把"这是超时被杀，不是崩溃/脚本报错"写清楚，并给出可行替代做法。
+  const loader = createTsModuleLoader({
+    mocks: {
+      "@tauri-apps/api/core": {
+        async invoke(command) {
+          assert.equal(command, "shell_run");
+          return {
+            exit_code: -1,
+            shell: "bash",
+            platform: "macos",
+            profile: "posix-bash",
+            shell_family: "posix",
+            stdout: "",
+            stderr: "",
+            stdout_truncated: false,
+            stderr_truncated: false,
+            timed_out: true,
+            cancelled: false,
+            effective_timeout_ms: 30_000,
+            duration_ms: 30_123,
+          };
+        },
+      },
+    },
+  });
+
+  const { createShellTools } = loader.loadModule("src/lib/tools/shellTools.ts");
+  const bundle = createShellTools({
+    workdir: "/repo",
+    providerId: "deepseek",
+    runtimePlatform: "macos",
+  });
+
+  const result = await bundle.executeToolCall(createBashCall("make build"));
+
+  assert.equal(result.isError, true);
+  assert.equal(result.details.timed_out, true);
+  const text = result.content[0].text;
+  assert.match(text, /timed_out: true/);
+  assert.match(text, /timeout_ms: 30000/);
+  assert.match(text, /killed by the shell timeout \(30000ms\)/);
+  assert.match(text, /it did not crash/);
+  assert.match(text, /DeepSeek caps Bash at 30000ms/);
+  assert.match(text, /re-running the identical command unchanged reaches the same limit/);
+  assert.match(text, /redirect the output to a file/);
+});