fix(agents): surface exec failures after claimed success

jbetala7 · jbetala7 · commit 29879808eb14 · 2026-05-10T04:41:35.000+05:30
diff --git a/src/agents/pi-embedded-runner/run/payloads.errors.test.ts b/src/agents/pi-embedded-runner/run/payloads.errors.test.ts
@@ -378,6 +378,23 @@ describe("buildEmbeddedRunPayloads", () => {
     expect(payloads[1]?.text).not.toContain("missing");
   });
 
+  it("shows exec tool errors when assistant output claims success", () => {
+    const payloads = buildPayloads({
+      assistantTexts: ["The script is ready to use and saved in your workspace."],
+      lastAssistant: { stopReason: "end_turn" } as unknown as AssistantMessage,
+      lastToolError: {
+        toolName: "exec",
+        error: "/bin/bash: line 1: python: command not found",
+      },
+    });
+
+    expect(payloads).toHaveLength(2);
+    expect(payloads[0]?.text).toBe("The script is ready to use and saved in your workspace.");
+    expect(payloads[1]?.isError).toBe(true);
+    expect(payloads[1]?.text).toContain("Exec");
+    expect(payloads[1]?.text).not.toContain("python: command not found");
+  });
+
   it("shows mutating tool errors when assistant output does not acknowledge the failure", () => {
     const payloads = buildPayloads({
       assistantTexts: ["No issues found. The update is complete."],
@@ -435,6 +452,17 @@ describe("buildEmbeddedRunPayloads", () => {
     expectSinglePayloadSummary(payloads, { text });
   });
 
+  it("suppresses exec warnings when assistant output explicitly acknowledges the command failure", () => {
+    const text = "I couldn't run the command because python was not found.";
+    const payloads = buildPayloads({
+      assistantTexts: [text],
+      lastAssistant: { stopReason: "end_turn" } as unknown as AssistantMessage,
+      lastToolError: { toolName: "exec", error: "/bin/bash: line 1: python: command not found" },
+    });
+
+    expectSinglePayloadSummary(payloads, { text });
+  });
+
   it("does not treat session_status read failures as mutating when explicitly flagged", () => {
     const payloads = buildPayloads({
       assistantTexts: ["Status loaded."],
diff --git a/src/agents/pi-embedded-runner/run/payloads.test.ts b/src/agents/pi-embedded-runner/run/payloads.test.ts
@@ -88,11 +88,28 @@ describe("buildEmbeddedRunPayloads tool-error warnings", () => {
     expectSinglePayloadText(payloads, "Fixed.");
   });
 
-  it("suppresses exec tool errors when verbose mode is off", () => {
-    expectNoPayloads({
+  it("surfaces concise exec tool errors when verbose mode is off", () => {
+    const payloads = buildPayloads({
       lastToolError: { toolName: "exec", error: "command failed" },
       verboseLevel: "off",
     });
+
+    expectSingleToolErrorPayload(payloads, {
+      title: "Exec",
+      absentDetail: "command failed",
+    });
+  });
+
+  it("surfaces concise bash tool errors when verbose mode is off", () => {
+    const payloads = buildPayloads({
+      lastToolError: { toolName: "bash", error: "command failed" },
+      verboseLevel: "off",
+    });
+
+    expectSingleToolErrorPayload(payloads, {
+      title: "Bash",
+      absentDetail: "command failed",
+    });
   });
 
   it("surfaces exec tool errors for cron sessions even when verbose mode is off", () => {
@@ -132,12 +149,17 @@ describe("buildEmbeddedRunPayloads tool-error warnings", () => {
     });
   });
 
-  it("keeps non-timeout exec tool errors suppressed for cron sessions when verbose mode is off", () => {
-    expectNoPayloads({
+  it("surfaces non-timeout exec tool errors for cron sessions without raw details", () => {
+    const payloads = buildPayloads({
       lastToolError: { toolName: "exec", error: "Command not found" },
       sessionKey: "agent:main:cron:job-1",
       verboseLevel: "off",
     });
+
+    expectSingleToolErrorPayload(payloads, {
+      title: "Exec",
+      absentDetail: "Command not found",
+    });
   });
 
   it("shows exec tool errors when verbose mode is on", () => {
diff --git a/src/agents/pi-embedded-runner/run/payloads.ts b/src/agents/pi-embedded-runner/run/payloads.ts
@@ -24,10 +24,7 @@ import {
   normalizeTextForComparison,
 } from "../../pi-embedded-helpers.js";
 import type { ToolResultFormat } from "../../pi-embedded-subscribe.shared-types.js";
-import {
-  extractAssistantThinking,
-  extractAssistantVisibleText,
-} from "../../pi-embedded-utils.js";
+import { extractAssistantThinking, extractAssistantVisibleText } from "../../pi-embedded-utils.js";
 import { isExecLikeToolName, type ToolErrorSummary } from "../../tool-error-summary.js";
 import { isLikelyMutatingToolName } from "../../tool-mutation.js";
 
@@ -48,7 +45,7 @@ const RECOVERABLE_TOOL_ERROR_KEYWORDS = [
 ] as const;
 
 const MUTATING_FAILURE_ACTION_PATTERN =
-  "(?:write|edit|update|save|create|delete|remove|modify|change|apply|patch|move|rename|send|reply|message|tool|action|operation)";
+  "(?:write|edit|update|save|create|delete|remove|modify|change|apply|patch|move|rename|send|reply|message|run|execute|execution|command|script|shell|bash|exec|tool|action|operation)";
 
 const MUTATING_FAILURE_INABILITY_PATTERN = new RegExp(
   `\\b(?:couldn't|could not|can't|cannot|unable to|am unable to|wasn't able to|was not able to|were unable to)\\b.{0,100}\\b${MUTATING_FAILURE_ACTION_PATTERN}\\b`,
@@ -143,9 +140,6 @@ function resolveToolErrorWarningPolicy(params: {
   if (params.suppressToolErrorWarnings) {
     return { showWarning: false, includeDetails };
   }
-  if (isExecLikeToolName(params.lastToolError.toolName) && !includeDetails) {
-    return { showWarning: false, includeDetails };
-  }
   // sessions_send timeouts and errors are transient inter-session communication
   // issues — the message may still have been delivered. Suppress warnings to
   // prevent raw error text from leaking into the chat surface (#23989).
@@ -160,6 +154,9 @@ function resolveToolErrorWarningPolicy(params: {
       includeDetails,
     };
   }
+  if (isExecLikeToolName(params.lastToolError.toolName) && !includeDetails) {
+    return { showWarning: false, includeDetails };
+  }
   if (params.suppressToolErrors) {
     return { showWarning: false, includeDetails };
   }
