Merge pull request #5885 from deepseek-harness/fix/team-send-tool-receipt

fix(agent-team): omit message IDs from send tool results
2026-10-08 16:10:36 +08:00
commit 12b8066c28
@@ -264,8 +264,8 @@
  en: f9457f033f75f00a
  zh: a2d60ec44118c34c
 /tool-schema-catalog/deepseek-ai-dsh-experimental-tool-agent-team/send-message:
-  en: 0746bb11dae8db55
-  zh: 8b634add76c4d88a
+  en: ec50e6a0b86aecc6
+  zh: 29c8728536a5adf1
 /tool-schema-catalog/deepseek-ai-dsh-experimental-tool-agent-team/spawn-teammate:
  en: 02d6cc746e234b99
  zh: 49bc60ac098bca48
@@ -2257,7 +2257,7 @@ Source: [`packages/experimental/tool-agent-team/src/index.ts`](../packages/exper

 ### `send_message`

-Send one message to another Team member and return its accepted message id. A running target receives it at the nearest step boundary; an inactive target starts or resumes a turn.
+Send one message to another Team member. A running target receives it at the nearest step boundary; an inactive target starts or resumes a turn.

 ```json
 {
@@ -2263,7 +2263,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后，

 ### `send_message`

-向另一名 Team member 发送一条消息，并返回获准消息 id。running target 会在最近的步骤边界收到消息；inactive target 会启动或恢复一个 turn。
+向另一名 Team member 发送一条消息。running target 会在最近的步骤边界收到消息；inactive target 会启动或恢复一个 turn。

 ```json
 {
@@ -9,8 +9,8 @@
  en: 04d7dcbdc1679eae
  zh: 8b083641a9262d80
 /agent-teams-messages-use-target-inboxes/change:
-  en: 94d62cf9d98270ab
-  zh: c589725cd35ccd06
+  en: 9968decf3dd340af
+  zh: 736a32caac4b5103
 /agent-teams-messages-use-target-inboxes/migration:
-  en: f7811928da727edc
-  zh: 9a28b6b745725a6d
+  en: 7ffd442bfd55769e
+  zh: ffb11a143f11ce7f
@@ -9,7 +9,7 @@ English | [中文](guide.zh.md)

 ## Change

-The experimental Agent Teams plugin sends new messages directly to the target Agent inbox instead of retaining a separate Team mailbox. `send_message` returns only `messageId`; it no longer returns an `accepted` or `queued` status. Acceptance does not confirm model processing or a synchronous storage flush. Failed sends are not retained for retry, and explicit resends are not deduplicated.
+The experimental Agent Teams plugin sends new messages directly to the target Agent inbox instead of retaining a separate Team mailbox. `send_message` returns only `{ "sent": true }`; it exposes neither a message ID nor an `accepted` or `queued` status. Acceptance does not confirm model processing or a synchronous storage flush. Failed sends are not retained for retry, and explicit resends are not deduplicated.

 The `maxPendingMessagesPerMember` configuration key is removed. Historical `team/message/queued` and `team/message/delivered` records remain readable, but queued messages that were not delivered before upgrading are never dispatched. Roster and task state remain durable.

@@ -17,5 +17,5 @@ The `maxPendingMessagesPerMember` configuration key is removed. Historical `team

 1. Before upgrading, let pending Team messages finish and confirm that their targets received them. After upgrading, explicitly resend any required undelivered message; check the target's history first to avoid duplicates. Do not edit historical Session records.
 2. Remove `maxPendingMessagesPerMember` from the `@deepseek-ai/dsh-experimental-agent-team` plugin configuration in your `cordis.yml` or overlays. There is no replacement Team queue limit; `maxMessageBytes` still bounds each complete sender-framed message.
-3. Update any consumer of `send_message` results to use `messageId` alone and handle send errors explicitly. Do not treat success as completed target work.
-4. Start your Team-enabled profile, send a message to another member, and confirm that the tool returns `messageId` and that the target receives it. Consult the [package reference](../../../../packages/experimental/agent-team/README.md#use-this-package) for delivery semantics.
+3. Update any consumer of `send_message` results to read `sent` instead of message IDs or delivery statuses, and handle send errors explicitly. Do not treat success as completed target work. The `ctx.agentTeams.sendMessage()` service still returns `messageId`.
+4. Start your Team-enabled profile, send a message to another member, and confirm that the tool returns `{ "sent": true }` and that the target receives it. Consult the [package reference](../../../../packages/experimental/agent-team/README.md#use-this-package) for delivery semantics.
@@ -9,7 +9,7 @@ description: "Agent Teams 移除待投递消息数量限制，并停止投递历

 ## 变更

-实验性 Agent Teams 插件将新消息直接发送到目标 Agent 收件箱，不再保留独立的 Team 邮箱。`send_message` 仅返回 `messageId`，不再返回 `accepted` 或 `queued` 状态。接收成功不表示模型已处理消息，也不保证同步刷新到存储。失败的发送不会保留以供重试，显式重发也不会去重。
+实验性 Agent Teams 插件将新消息直接发送到目标 Agent 收件箱，不再保留独立的 Team 邮箱。`send_message` 仅返回 `{ "sent": true }`，不暴露消息 ID，也不返回 `accepted` 或 `queued` 状态。接收成功不表示模型已处理消息，也不保证同步刷新到存储。失败的发送不会保留以供重试，显式重发也不会去重。

 配置键 `maxPendingMessagesPerMember` 已移除。历史 `team/message/queued` 和 `team/message/delivered` 记录仍可读取，但升级前尚未投递的排队消息永远不会被派发。成员名单与任务状态仍然持久保存。

@@ -17,5 +17,5 @@ description: "Agent Teams 移除待投递消息数量限制，并停止投递历

 1. 升级前，让待投递的 Team 消息完成投递，并确认目标已收到。升级后，显式重发仍需处理的未投递消息；先检查目标历史，避免重复。不要编辑历史 Session 记录。
 2. 从 `cordis.yml` 或覆盖配置中 `@deepseek-ai/dsh-experimental-agent-team` 插件的配置移除 `maxPendingMessagesPerMember`。没有替代的 Team 队列限制；`maxMessageBytes` 仍限制包含发送者标识的完整消息大小。
-3. 更新所有读取 `send_message` 结果的调用方，仅使用 `messageId`，并显式处理发送错误。不要把发送成功视为目标工作已完成。
-4. 启动启用 Team 的 profile，向其他成员发送消息，确认工具返回 `messageId` 且目标收到消息。投递语义参见[包参考](../../../../packages/experimental/agent-team/README.zh.md#use-this-package)。
+3. 更新所有读取 `send_message` 结果的调用方，改为读取 `sent` 而非消息 ID 或投递状态，并显式处理发送错误。不要把发送成功视为目标工作已完成。`ctx.agentTeams.sendMessage()` 服务仍返回 `messageId`。
+4. 启动启用 Team 的 profile，向其他成员发送消息，确认工具返回 `{ "sent": true }` 且目标收到消息。投递语义参见[包参考](../../../../packages/experimental/agent-team/README.zh.md#use-this-package)。
@@ -27,8 +27,8 @@
  en: be42811bbcc3b438
  zh: 014bd055ca694315
 /deepseek-ai-dsh-experimental-tool-agent-team/use-this-package/what-success-and-failure-look-like:
-  en: 259e9d39e1d1579a
-  zh: c61feed86e1ba2d2
+  en: e49a980372c55892
+  zh: d4bcdd36278d040d
 /deepseek-ai-dsh-experimental-tool-agent-team/understand-the-implementation:
  en: 3952eb1d9de1f988
  zh: 789d51b88f41b710
@@ -65,7 +65,7 @@ Creation and listing results identify members by `target`, with no member Sessio

 ### What success and failure look like

-Sending a message succeeds as soon as it is safely stored: the result is `accepted` (delivered now) or `queued` (waiting), and a queued message must not be resent. `wait_agent` returns `noProgress` right away when no other member is running or provisioning, telling the caller to wake a teammate first; otherwise it waits for the next change and the caller re-reads state afterward. Task edits based on an outdated revision are rejected rather than overwriting newer work.
+`send_message` returns `{ "sent": true }` after inbox acceptance, without exposing the message ID. Success does not confirm recipient processing or a synchronous storage flush; send failures are reported as tool errors. `wait_agent` returns `noProgress` right away when no other member is running or provisioning, telling the caller to wake a teammate first; otherwise it waits for the next change and the caller re-reads state afterward. Task edits based on an outdated revision are rejected rather than overwriting newer work.

 -----

@@ -65,7 +65,7 @@ kind: "package-reference"

 ### 成功与失败的表现

-发送消息在安全存储后即成功：结果为 `accepted`（已立即送达）或 `queued`（等待中），排队的消息绝不能重发。当没有其他成员 running 或 provisioning 时，`wait_agent` 会立即返回 `noProgress`，提示调用方先唤醒 teammate；否则它会等待下一次变化，调用方随后重新读取状态。基于过期 revision 的任务编辑会被拒绝，而不是覆盖更新的成果。
+`send_message` 在收件箱接收消息后返回 `{ "sent": true }`，不暴露消息 ID。成功不表示接收方已处理消息，也不保证同步刷新到存储；发送失败会作为工具错误报告。当没有其他成员 running 或 provisioning 时，`wait_agent` 会立即返回 `noProgress`，提示调用方先唤醒 teammate；否则它会等待下一次变化，调用方随后重新读取状态。基于过期 revision 的任务编辑会被拒绝，而不是覆盖更新的成果。

 -----

@@ -34,7 +34,7 @@ The Team Lead and all teammates share the same working directory and filesystem.

 Prefer read/edit/write for file changes. If a file operation returns FS_STALE_VERSION, read the current file, rebase your intended change onto the new content, and retry. Bash, formatters, code generators, and scripts are not fully protected by the filesystem version guard; coordinate them explicitly and have the Lead review the final diff and run tests.

-Use the target returned by spawn_teammate or list_agents for send_message and interrupt_agent, or as owner when assigning or filtering shared tasks. send_message steers a running target at its nearest step boundary and starts or resumes an inactive target. inactive means no turn is executing; it does not describe task completion, success, failure, or waiting for other agents. provisioning means member creation is in progress; failed means member creation failed. A peer message starts with its sender name. A successful send returns the accepted message id; it does not mean the recipient has processed it. Shared-task workflow is list, get, claim with the current revision, perform the work, then complete. Task readiness never starts an owner. Before wait_agent, use list_agents and make sure another required member is running or provisioning; use send_message first when the required member is inactive. wait_agent observes only changes after that call starts, never wakes a member, and returns noProgress immediately when no other member can produce a change. Re-list after wakeup or timeout. The Lead must wait for required teammates before giving the final answer.`
+Use the target returned by spawn_teammate or list_agents for send_message and interrupt_agent, or as owner when assigning or filtering shared tasks. send_message steers a running target at its nearest step boundary and starts or resumes an inactive target. inactive means no turn is executing; it does not describe task completion, success, failure, or waiting for other agents. provisioning means member creation is in progress; failed means member creation failed. A peer message starts with its sender name. A successful send does not mean the recipient has processed it. Shared-task workflow is list, get, claim with the current revision, perform the work, then complete. Task readiness never starts an owner. Before wait_agent, use list_agents and make sure another required member is running or provisioning; use send_message first when the required member is inactive. wait_agent observes only changes after that call starts, never wakes a member, and returns noProgress immediately when no other member can produce a change. Re-list after wakeup or timeout. The Lead must wait for required teammates before giving the final answer.`

 const ACTIVE_WAIT_STATUSES: ReadonlySet<TeamMemberView['status']> = new Set(['running', 'provisioning'])
 const NO_ACTIVE_PEER_MESSAGE = 'No other Team member is running or provisioning. wait_agent cannot make progress or wake inactive teammates. Re-list with list_agents and team_task_list, then use send_message to wake each required inactive teammate before waiting again.'
@@ -97,7 +97,7 @@ const SEND_VALUE_SCHEMA = {
  type: 'object',
  additionalProperties: false,
  properties: {
-    messageId: { type: 'string', required: true },
+    sent: { type: 'boolean', required: true, const: true },
  },
 } as const

@@ -213,18 +213,19 @@ To message another teammate, use send_message({ target: "<teammate name>", messa

    register(scoped.tools.register(defineTool({
      name: 'send_message',
-      description: 'Send one message to another Team member and return its accepted message id. A running target receives it at the nearest step boundary; an inactive target starts or resumes a turn.',
+      description: 'Send one message to another Team member. A running target receives it at the nearest step boundary; an inactive target starts or resumes a turn.',
      parameters: {
        target: { type: 'string', required: true, description: 'Member target returned by spawn_teammate or list_agents, including lead.' },
        message: { type: 'string', required: true, description: 'Self-contained message for the target.' },
      },
      output: jsonOutput(SEND_VALUE_SCHEMA),
-      execute(args, exec) {
-        return ctx.agentTeams.sendMessage(callingAgent(exec.agent, 'send_message'), {
+      async execute(args, exec) {
+        await ctx.agentTeams.sendMessage(callingAgent(exec.agent, 'send_message'), {
          target: args.target,
          content: [{ type: 'text', text: args.message }],
          signal: exec.signal,
        })
+        return { sent: true }
      },
    })))

@@ -202,6 +202,7 @@ describe('dsh-tool-team', () => {
    expect(JSON.parse(text(tasks))).toMatchObject({ tasks: [{ id: task.id, ownerName: member.target }] })
    const sent = await execute(ctx, lead, 'send_message', { target: member.target, message: 'review the diff' })
    expect(sent.isError).toBe(false)
+    expect(JSON.parse(text(sent))).toEqual({ sent: true })
    const interrupted = await execute(ctx, lead, 'interrupt_agent', { target: listed[1]!.target })
    expect(interrupted.isError).toBe(false)
    await child.whenIdle()
@@ -483,10 +484,10 @@ describe('dsh-tool-team', () => {
    expect(text(roster)).toBe(JSON.stringify(JSON.parse(text(roster))))
    const peer = await execute(ctx, child, 'send_message', { target: 'lead', message: 'progress report' })
    expect(peer.isError).toBe(false)
-    expect(JSON.parse(text(peer))).toHaveProperty('messageId')
+    expect(JSON.parse(text(peer))).toEqual({ sent: true })
    const followup = await execute(ctx, child, 'send_message', { target: 'lead', message: 'review the report' })
    expect(followup.isError).toBe(false)
-    expect(JSON.parse(text(followup))).toHaveProperty('messageId')
+    expect(JSON.parse(text(followup))).toEqual({ sent: true })
    await lead.whenIdle()

    const created = await execute(ctx, lead, 'team_task_create', {
@@ -27,7 +27,7 @@
 {"type":"step/start","data":{"turn":1,"step":3}}
 {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"tool-call","id":"direct-3","name":"send_message","arguments":"{\"target\":\"reviewer\",\"message\":\"Review the direct inbox delivery.\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1790429366017,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790429366017,"index":0,"dt":[],"id":"direct-3","name":"send_message","args":["{\"target\":\"reviewer\",\"message\":\"Review the direct inbox delivery.\"}"]},{"type":"chunk","time":1790429366017,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"direct-3","name":"send_message","arguments":"{\"target\":\"reviewer\",\"message\":\"Review the direct inbox delivery.\"}"}}},{"type":"chunk","time":1790429366017,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1790429366017,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}
 {"type":"tool/call","data":{"turn":1,"step":3,"callId":"direct-3","name":"send_message","arguments":"{\"target\":\"reviewer\",\"message\":\"Review the direct inbox delivery.\"}"}}
-{"type":"tool/result","data":{"turn":1,"step":3,"message":{"role":"tool","source":{"kind":"tool","callId":"direct-3"},"toolCallId":"direct-3","content":[{"type":"text","text":"{\"messageId\":\"{{message:17}}\"}"}],"isError":false,"id":"{{message:9}}"}},"sourceEventSeqs":[27],"surfaceOp":"append"}
+{"type":"tool/result","data":{"turn":1,"step":3,"message":{"role":"tool","source":{"kind":"tool","callId":"direct-3"},"toolCallId":"direct-3","content":[{"type":"text","text":"{\"sent\":true}"}],"isError":false,"id":"{{message:9}}"}},"sourceEventSeqs":[27],"surfaceOp":"append"}
 {"type":"step/end","data":{"turn":1,"step":3}}
 {"type":"step/start","data":{"turn":1,"step":4}}
 {"type":"assistant/message","data":{"turn":1,"step":4,"message":{"role":"assistant","content":[{"type":"tool-call","id":"direct-4","name":"interrupt_agent","arguments":"{\"target\":\"reviewer\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:10}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1790429366033,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790429366033,"index":0,"dt":[],"id":"direct-4","name":"interrupt_agent","args":["{\"target\":\"reviewer\"}"]},{"type":"chunk","time":1790429366033,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"direct-4","name":"interrupt_agent","arguments":"{\"target\":\"reviewer\"}"}}},{"type":"chunk","time":1790429366033,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1790429366033,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}
@@ -11,7 +11,7 @@ The Team Lead and all teammates share the same working directory and filesystem.

 Prefer read/edit/write for file changes. If a file operation returns FS_STALE_VERSION, read the current file, rebase your intended change onto the new content, and retry. Bash, formatters, code generators, and scripts are not fully protected by the filesystem version guard; coordinate them explicitly and have the Lead review the final diff and run tests.

-Use the target returned by spawn_teammate or list_agents for send_message and interrupt_agent, or as owner when assigning or filtering shared tasks. send_message steers a running target at its nearest step boundary and starts or resumes an inactive target. inactive means no turn is executing; it does not describe task completion, success, failure, or waiting for other agents. provisioning means member creation is in progress; failed means member creation failed. A peer message starts with its sender name. A successful send returns the accepted message id; it does not mean the recipient has processed it. Shared-task workflow is list, get, claim with the current revision, perform the work, then complete. Task readiness never starts an owner. Before wait_agent, use list_agents and make sure another required member is running or provisioning; use send_message first when the required member is inactive. wait_agent observes only changes after that call starts, never wakes a member, and returns noProgress immediately when no other member can produce a change. Re-list after wakeup or timeout. The Lead must wait for required teammates before giving the final answer.
+Use the target returned by spawn_teammate or list_agents for send_message and interrupt_agent, or as owner when assigning or filtering shared tasks. send_message steers a running target at its nearest step boundary and starts or resumes an inactive target. inactive means no turn is executing; it does not describe task completion, success, failure, or waiting for other agents. provisioning means member creation is in progress; failed means member creation failed. A peer message starts with its sender name. A successful send does not mean the recipient has processed it. Shared-task workflow is list, get, claim with the current revision, perform the work, then complete. Task readiness never starts an owner. Before wait_agent, use list_agents and make sure another required member is running or provisioning; use send_message first when the required member is inactive. wait_agent observes only changes after that call starts, never wakes a member, and returns noProgress immediately when no other member can produce a change. Re-list after wakeup or timeout. The Lead must wait for required teammates before giving the final answer.

 Check the [exit code: N] marker on every bash result; investigate failures before moving on.

@@ -293,7 +293,7 @@
    },
    {
      "name": "send_message",
-      "description": "Send one message to another Team member and return its accepted message id. A running target receives it at the nearest step boundary; an inactive target starts or resumes a turn.",
+      "description": "Send one message to another Team member. A running target receives it at the nearest step boundary; an inactive target starts or resumes a turn.",
      "parameters": {
        "type": "object",
        "properties": {