feat(native): every post names the message the turn answers; hosted prompt says it cannot change a task's status

backend/__tests__/unit/services/nativeRuntimeService.claims.test.js

@@ -251,4 +251,59 @@ describe('nativeRuntimeService claim-before-act (ADR-018 D3)', () => {
     await runAgent(installation, { type: 'heartbeat', eventId: 'evt-2', payload: {} });
     expect(MessageClaimService.claim).not.toHaveBeenCalled();
   });
+
+  // Walk 5 (2026-10-16): beaten to the claim, the addressed agent answered in
+  // 13 s and TASK-002 never completed, because the answer was credited only
+  // through a live claim. Every post the turn makes now names the message it
+  // answers, taken from the event.
+  describe('every post names the message the turn answers', () => {
+    const withPostTool = { ...installation, config: { ...installation.config, tools: ['commonly_post_message'] } };
+    const postCall = (args) => ({
+      status: 200,
+      headers: {},
+      data: {
+        choices: [{ message: { role: 'assistant', content: null, tool_calls: [{ id: 'c1', type: 'function', function: { name: 'commonly_post_message', arguments: JSON.stringify(args) } }] } }],
+        usage: { prompt_tokens: 10, completion_tokens: 5, total_tokens: 15 },
+      },
+    });
+    const done = {
+      status: 200, headers: {}, data: { choices: [{ message: { role: 'assistant', content: 'NO_REPLY' } }], usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 } },
+    };
+
+    test('beaten to the claim, the tool post still carries the trigger id; the model cannot set it', async () => {
+      MessageClaimService.claim.mockResolvedValue({ claimed: false, claimedBy: 'scout', instanceId: 'scout-1' });
+      mockHolderEvent.mockResolvedValue(null);
+      axios.post.mockResolvedValueOnce(postCall({ content: 'Here is the answer.', answeringMessageId: 'msg-999' })).mockResolvedValueOnce(done);
+
+      await runAgent(withPostTool, mentionTrigger);
+
+      expect(AgentMessageService.postMessage).toHaveBeenCalledTimes(1);
+      expect(AgentMessageService.postMessage).toHaveBeenCalledWith(expect.objectContaining({
+        content: 'Here is the answer.', answeringMessageId: 'msg-77',
+      }));
+    });
+
+    test('the text fallback post carries it too', async () => {
+      axios.post.mockResolvedValueOnce({
+        status: 200, headers: {}, data: { choices: [{ message: { role: 'assistant', content: 'Plain answer.' } }], usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 } },
+      });
+
+      await runAgent(installation, mentionTrigger);
+
+      expect(AgentMessageService.postMessage).toHaveBeenCalledWith(expect.objectContaining({
+        content: 'Plain answer.', metadata: expect.objectContaining({ fallback: true }), answeringMessageId: 'msg-77',
+      }));
+    });
+
+    test('a turn no message woke answers no message', async () => {
+      axios.post.mockResolvedValueOnce({
+        status: 200, headers: {}, data: { choices: [{ message: { role: 'assistant', content: 'Heartbeat note.' } }], usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 } },
+      });
+
+      await runAgent(installation, { type: 'heartbeat', eventId: 'evt-3', payload: {} });
+
+      expect(AgentMessageService.postMessage).toHaveBeenCalledWith(expect.objectContaining({ content: 'Heartbeat note.' }));
+      expect(AgentMessageService.postMessage.mock.calls[0][0].answeringMessageId).toBeUndefined();
+    });
+  });
 });

backend/__tests__/unit/services/nativeRuntimeService.hostedEngine.test.js

@@ -316,6 +316,8 @@ describe('nativeRuntimeService — hosted installs on the native engine', () =>
         status: 'open', total: 3, tasks: [{ taskId: 'TASK-1', title: 'Say hello to the pod', status: 'pending', assignee: null }],
       });
       expect(systemPromptIn(0)).toMatch(/check the board with commonly_list_tasks before saying what is or is not on it/);
+      // Walk 5: it offered to "mark TASK-002 done" with no tool that can.
+      expect(systemPromptIn(0)).toMatch(/you cannot change a task's status or mark one done: if asked to, say so/);
       // QA walk: a pod summary also searched the web and ran code (19 s).
       expect(systemPromptIn(0)).toMatch(/Answer a question about this pod from the pod itself; reach for the tools below only when/);
       // QA walk: an answer opening "@asker" filed every reply as "needs you".

backend/__tests__/unit/services/nativeRuntimeService.hostedOffer.test.js

@@ -98,6 +98,16 @@ describe('commonly_offer_hosted_agent', () => {
     expect(posted.payload.starts[0].href).toMatch(/^\/v2\/[^/\\]/);
   });
 
+  test('the offer names the message it answers, from the event', async () => {
+    axios.post
+      .mockResolvedValueOnce(callsTool('commonly_offer_hosted_agent', { content: REPLY }))
+      .mockResolvedValueOnce(says('Posted the button above.'));
+
+    await runAgent(scoutInstall, { ...ask, payload: { ...ask.payload, messageId: '5001' } });
+
+    expect(AgentMessageService.postMessage.mock.calls[0][0].answeringMessageId).toBe('5001');
+  });
+
   test('posts nothing when the person already holds the beta\'s hosted agent, and names it', async () => {
     mockListHosted.mockResolvedValue([{ agentName: 'lee-agent', instanceId: 'default', podId: POD }]);
     axios.post

backend/services/hostedEngine.ts

@@ -223,9 +223,13 @@ export const hostedNativeSystemPrompt = (
     + 'when the mention alone is not '
     + 'enough. Keep durable facts with commonly_write_memory, and track work with '
     + 'commonly_create_task; check the board with commonly_list_tasks before saying what is or is '
+    // Walk 5 (2026-10-16): the agent offered to "mark TASK-002 done", which
+    // no hosted tool can do.
+    + 'not on it. You can add tasks and read the board, but you cannot change a task\'s status or '
+    + 'mark one done: if asked to, say so and point them to the pod\'s board. '
     // QA walk 2026-10-16: asked to summarize the pod, a hosted agent also
     // searched the web and ran code, and the reply took 19 s instead of 7.
-    + 'not on it. Answer a question about this pod from the pod itself; reach for the tools below '
+    + 'Answer a question about this pod from the pod itself; reach for the tools below '
     + `only when the answer needs something outside it. ${uses.join(' ')} If no reply is genuinely `
     + 'needed, do not post.\n'
     + '\n'

backend/services/nativeRuntimeService.ts

@@ -506,6 +506,14 @@ interface DispatchContext {
    * resolved, and null when it cannot be: then no pod message gives an address.
    */
   ownUserId: string | null;
+  /**
+   * The message this turn answers: the trigger's own messageId, from the
+   * event, never from the model. Every post the turn makes carries it so the
+   * answer is credited to that message's author even when the turn runs
+   * unclaimed (an addressed agent that lost the claim race, #2269, or a claim
+   * service that failed open). It is not a reply edge and threads nothing.
+   */
+  answeringMessageId?: string;
 }
 
 // The agent's own bot User, found the way the posting path derives it (the
@@ -644,6 +652,7 @@ async function dispatchTool(
           // Web results are in this turn: the post stays in the pod and never
           // reaches a channel relay, whatever bridges exist when it is stored.
           noOutboundRelay: Boolean(ctx.webTainted),
+          answeringMessageId: ctx.answeringMessageId,
         });
         return { content: { ok: true, posted: true, result } };
       }
@@ -688,6 +697,7 @@ async function dispatchTool(
           metadata: { source: 'native-runtime' },
           payload: { kind: 'starts', starts: [{ id: 'run-hosted', label: HOSTED_OFFER_LABEL, href }] },
           noOutboundRelay: Boolean(ctx.webTainted),
+          answeringMessageId: ctx.answeringMessageId,
         });
         return { content: { ok: true, posted: true, result } };
       }
@@ -1538,6 +1548,7 @@ export async function runAgent(
     installationConfig: cfg,
     ownUserId: null,
     givenAddresses: new GivenAddresses(),
+    answeringMessageId: claimMessageId || undefined,
   };
   // Who wrote what matters only to commonly_fetch_url, so a turn that cannot
   // fetch makes neither lookup (its own user, the trigger's sender).
@@ -1915,6 +1926,7 @@ export async function runAgent(
             installationConfig: cfg,
             metadata: { source: 'native-runtime', fallback: true },
             noOutboundRelay: Boolean(dispatchCtx.webTainted),
+            answeringMessageId: dispatchCtx.answeringMessageId,
           });
           // The message service suppresses a bare NO_REPLY (and other
           // silent output) without throwing. That is a normal declined turn,