diff --git a/src/core/bot.ts b/src/core/bot.ts
index 3312b33..3158d91 100644
--- a/src/core/bot.ts
+++ b/src/core/bot.ts
@@ -1222,6 +1222,7 @@ export class LettaBot implements AgentSession {
       const maxRepeatedBashFailures = 3;
       let lastEventType: string | null = null;
       let abortedWithMessage = false;
+      let turnError: string | undefined;
 
       const parseAndHandleDirectives = async () => {
         if (!response.trim()) return;
@@ -1346,6 +1347,7 @@ export class LettaBot implements AgentSession {
                 log.error(`Agent stuck in tool loop (${msgTypeCounts['tool_call']} calls), aborting`);
                 session?.abort().catch(() => {});
                 response = '(Agent got stuck in a tool loop and was stopped. Try sending your message again.)';
+                turnError = `tool loop abort after ${msgTypeCounts['tool_call'] || 0} tool calls`;
                 abortedWithMessage = true;
                 break;
               }
@@ -1402,6 +1404,7 @@ export class LettaBot implements AgentSession {
                   log.error(`Stopping run after repeated Bash command failures (${repeatedBashFailureCount}) for: ${bashCommand}`);
                   session?.abort().catch(() => {});
                   response = `(I stopped after repeated CLI command failures while running: ${bashCommand}. The command path appears mismatched. Please confirm Bluesky CLI commands are available, then resend your request.)`;
+                  turnError = `repeated Bash failure abort (${repeatedBashFailureCount}x): ${bashCommand}`;
                   abortedWithMessage = true;
                   break;
                 }
@@ -1652,7 +1655,7 @@ export class LettaBot implements AgentSession {
             events,
             output: output || response,
             durationMs: Math.round(performance.now() - t0),
-            error: lastErrorDetail?.message,
+            error: turnError ?? lastErrorDetail?.message,
           }).catch(() => {});
         }
       }
diff --git a/src/core/result-guard.test.ts b/src/core/result-guard.test.ts
index 015d33e..ce9bb5e 100644
--- a/src/core/result-guard.test.ts
+++ b/src/core/result-guard.test.ts
@@ -111,6 +111,8 @@ describe('result divergence guard', () => {
       allowedTools: [],
       maxToolCalls: 100,
     });
+    const writeTurn = vi.fn(async () => {});
+    (bot as any).turnLogger = { write: writeTurn };
 
     const abort = vi.fn(async () => {});
     const adapter = {
@@ -152,6 +154,10 @@ describe('result divergence guard', () => {
     expect(abort).toHaveBeenCalled();
     const sentTexts = adapter.sendMessage.mock.calls.map(([payload]) => payload.text as string);
     expect(sentTexts.some(text => text.includes('repeated CLI command failures'))).toBe(true);
+    expect(writeTurn).toHaveBeenCalledTimes(1);
+    expect(writeTurn).toHaveBeenCalledWith(expect.objectContaining({
+      error: 'repeated Bash failure abort (3x): lettabot bluesky post --text "hi" --agent Bot',
+    }));
   });
 
   it('stops consuming stream and avoids retry after explicit tool-loop abort', async () => {
@@ -160,6 +166,8 @@ describe('result divergence guard', () => {
       allowedTools: [],
       maxToolCalls: 1,
     });
+    const writeTurn = vi.fn(async () => {});
+    (bot as any).turnLogger = { write: writeTurn };
 
     const adapter = {
       id: 'mock',
@@ -207,6 +215,10 @@ describe('result divergence guard', () => {
     expect(runSession).toHaveBeenCalledTimes(1);
     const sentTexts = adapter.sendMessage.mock.calls.map(([payload]) => payload.text);
     expect(sentTexts).toEqual(['(Agent got stuck in a tool loop and was stopped. Try sending your message again.)']);
+    expect(writeTurn).toHaveBeenCalledTimes(1);
+    expect(writeTurn).toHaveBeenCalledWith(expect.objectContaining({
+      error: 'tool loop abort after 1 tool calls',
+    }));
   });
 
   it('does not deliver reasoning text from error results as the response', async () => {