tianshu-mcp 0.5.6 → 0.5.7
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.en.md +16 -0
- package/CHANGELOG.md +16 -0
- package/README.en.md +25 -2
- package/README.md +26 -3
- package/dist/version.generated.js +1 -1
- package/package.json +1 -1
- package/skills/tianshu-mcp/SKILL.md +245 -113
- package/skills/tianshu-mcp/usage-examples.md +247 -159
package/CHANGELOG.en.md
CHANGED
|
@@ -8,6 +8,21 @@ Chinese version: [CHANGELOG.md](CHANGELOG.md)
|
|
|
8
8
|
|
|
9
9
|
---
|
|
10
10
|
|
|
11
|
+
## [0.5.7] - 2026-09-22
|
|
12
|
+
|
|
13
|
+
### Changed
|
|
14
|
+
|
|
15
|
+
- **The orchestration skill docs (`skills/tianshu-mcp/`) were rewritten item by item against the current code** (`SKILL.md` + `usage-examples.md`; the server idempotently syncs them into `~/.rivet/skills/tianshu-mcp/`). Every claim was checked against `src/mcp/tools.ts`, `src/config/schema.ts`, `src/tasks/task.ts`, `src/tasks/task-manager.ts`, `src/loop/fix-loop.ts`, `src/mcp/formatter.ts`, `src/agents/builtin.ts` and the five adapters:
|
|
16
|
+
- Added a **parameter compatibility matrix** (projectPath / model / modelSource / reasoningLevel / mode / planDoc / designSystem / allowCreateProject / continue_task across five agents) that states plainly what is rejected rather than silently ignored.
|
|
17
|
+
- Corrected the **Kimi Code tier domain**: it previously read `Low`/`High`/`Max`, but only `低/low`, `高/high`, `max`, `on` and `off` are accepted (deliberately no `中`/`medium`); the tier set is validated against the labels the UI actually renders, and an omitted tier forces `on` for unofficial models.
|
|
18
|
+
- Corrected the **`autoVerify` default** (on by default at the server level, previously undocumented) and completed the **`autoFixRounds` defaults** (codex 5 / zcode 2 / kimicode 2 / qoder 3 / traework falls back to the server default 0) plus the full `taskTimeoutMs` precedence.
|
|
19
|
+
- Added a full **qoder** section: `modelSource` disambiguation, saving and reading back the tier in Model Management, the global preference never being restored, send/answer checkpoints that never resend, automatic and manual repair both writing a plan and sending its full text back, macOS dispatch being disabled, and a multi-question JSON answer example.
|
|
20
|
+
- Added a **`needsUserKind` × agent × `continue_task` behaviour matrix** (six wait kinds × four agents), which previously was scattered per agent and never stated which kinds are confirmation-only and which re-send the complete brief.
|
|
21
|
+
- Added an **`agentEndReason` → terminal-state mapping** (`task_timeout` / `idle_timeout` / `cdp_disconnected` versus other hard failures landing in `needs_attention` or `failed(spawn)`), plus the missing `project_not_registered`, `unsupported_platform` and `qoder_error` (18 concrete codes) entries.
|
|
22
|
+
- Completed the **meta field table** with `qoderSessionId`, `actualModel`, `actualReasoningLevel`, `modelSource` and `guiStop`, and noted that `reasoningLevel` is an input that is never echoed while codex's effective tier is only visible in its panel.
|
|
23
|
+
- Clarified the three **`verify_task`** modes (task re-verification only updates the verdict fields; an independent `projectPath` accepts only a git ref as `baselineRef`; manual verification allocates a new report round) and the required arguments of `prepare_visual_baseline` / `approve_visual_baseline` (UUID plus a 64-hex digest).
|
|
24
|
+
- Removed duplicated material and split responsibilities: the main file teaches the method, the sub-file gives copy-ready shapes with wording matched to the real error strings.
|
|
25
|
+
|
|
11
26
|
## [0.5.6] - 2026-09-22
|
|
12
27
|
|
|
13
28
|
### Added
|
|
@@ -814,6 +829,7 @@ project → pick model and reasoning level → send instructions → run detecti
|
|
|
814
829
|
|
|
815
830
|
---
|
|
816
831
|
|
|
832
|
+
[0.5.7]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.6...v0.5.7
|
|
817
833
|
[0.5.6]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.5...v0.5.6
|
|
818
834
|
[0.5.5]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.4...v0.5.5
|
|
819
835
|
[0.5.4]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.3...v0.5.4
|
package/CHANGELOG.md
CHANGED
|
@@ -7,6 +7,21 @@
|
|
|
7
7
|
|
|
8
8
|
---
|
|
9
9
|
|
|
10
|
+
## [0.5.7] - 2026-09-22
|
|
11
|
+
|
|
12
|
+
### 变更
|
|
13
|
+
|
|
14
|
+
- **编排技能文档(`skills/tianshu-mcp/`)按当前代码逐项重写**(`SKILL.md` + `usage-examples.md`,server 启动时幂等同步到 `~/.rivet/skills/tianshu-mcp/`)。逐项核对了 `src/mcp/tools.ts`、`src/config/schema.ts`、`src/tasks/task.ts`、`src/tasks/task-manager.ts`、`src/loop/fix-loop.ts`、`src/mcp/formatter.ts`、`src/agents/builtin.ts` 与五个适配器的实际实现后修正偏差:
|
|
15
|
+
- 新增**参数兼容矩阵**(projectPath / model / modelSource / reasoningLevel / mode / planDoc / designSystem / allowCreateProject / continue_task 九个维度 × 五个 agent),明确「传错即报错」的边界。
|
|
16
|
+
- 修正 **Kimi Code 思考档位取值域**:此前写作 `Low`/`High`/`Max`,实际仅接受 `低/low`、`高/high`、`max`、`on`、`off`(刻意不含 `中`/`medium`),且档位集合按界面实际渲染标签校验、非官方模型不传时强制 `on`。
|
|
17
|
+
- 修正 **`autoVerify` 默认值**(server 默认开启,此前文档未说明),补全 **`autoFixRounds` 缺省轮数**(codex 5 / zcode 2 / kimicode 2 / qoder 3 / traework 落 server 默认 0)与 `taskTimeoutMs` 的完整优先级。
|
|
18
|
+
- 补齐 **qoder** 章节:`modelSource` 消歧、模型管理档位保存回读、全局偏好不还原、发送/答题检查点不重发、自动与手动返修均先落计划并回发全文、macOS 禁止派发;新增 qoder 多题续答 JSON 对象示例。
|
|
19
|
+
- 新增 **`needsUserKind` × agent × `continue_task` 行为矩阵**(六类等待 × 四类 agent),此前按 agent 零散描述且未写清哪些类型只作「已处理」确认、哪些会补发完整任务书。
|
|
20
|
+
- 新增 **`agentEndReason` → 终态映射表**(`task_timeout`/`idle_timeout`/`cdp_disconnected` 与其余硬失败分别落 `needs_attention` 或 `failed(spawn)`),并补 `project_not_registered`、`unsupported_platform`、`qoder_error`(含 18 个具体码)等缺失错误码。
|
|
21
|
+
- 补齐 **meta 字段全表**缺失项:`qoderSessionId`、`actualModel`、`actualReasoningLevel`、`modelSource`、`guiStop`(并说明 `reasoningLevel` 是入参不回显、codex 实际等级需看面板)。
|
|
22
|
+
- 明确 **`verify_task` 三种用法差异**(任务复验只更新结论字段;独立 projectPath 的 `baselineRef` 只能是 git ref;手动验收报告轮次分配)与 `prepare_visual_baseline`/`approve_visual_baseline` 的必填参数约束(UUID + 64 位十六进制摘要)。
|
|
23
|
+
- 清理重复与冗余内容,改为「主文件讲方法论 + 子文件给可复制形状」的分工;中英文表述与代码实际错误文案对齐。
|
|
24
|
+
|
|
10
25
|
## [0.5.6] - 2026-09-22
|
|
11
26
|
|
|
12
27
|
### 新增
|
|
@@ -709,6 +724,7 @@ Codex 桌面端改为 **GUI 驱动**:新增 `codex-gui` adapter,通过 MSIX
|
|
|
709
724
|
|
|
710
725
|
---
|
|
711
726
|
|
|
727
|
+
[0.5.7]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.6...v0.5.7
|
|
712
728
|
[0.5.6]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.5...v0.5.6
|
|
713
729
|
[0.5.5]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.4...v0.5.5
|
|
714
730
|
[0.5.4]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.5.3...v0.5.4
|
package/README.en.md
CHANGED
|
@@ -214,6 +214,10 @@ Use `server.log` when troubleshooting connections; do not treat stderr output it
|
|
|
214
214
|
| [docs/kimi-cdp.en.md](docs/kimi-cdp.en.md) | Kimi Code GUI driver: two renderer processes (main window + `Kimi Browser Overlay`), full-path workspace binding with native-dialog import, three-stage model selection and thinking tiers, execution mode, run detection and troubleshooting |
|
|
215
215
|
| [docs/codex-windows-smoke.en.md](docs/codex-windows-smoke.en.md) | Codex Windows hardware record (incl. verify-fail → auto plan → repair-pass loop) |
|
|
216
216
|
| [docs/release-v0.3.4.en.md](<docs/release-v0.3.4.en.md>) | v0.3.4 release notes (ZCode project/model read-back, initialization recovery, session dispatch confirmation, issues #8/#9/#10) |
|
|
217
|
+
| [docs/qoder-cdp.en.md](docs/qoder-cdp.en.md) | Qoder CN GUI driver: installation discovery and instance reuse, full-path workspaces with native import, `modelSource` and global Model Management reasoning tiers, send/answer checkpoints, liveness judging and same-session repair, hardware evidence and uncovered items |
|
|
218
|
+
| [docs/release-v0.5.7.en.md](<docs/release-v0.5.7.en.md>) | v0.5.7 release notes (orchestration skill docs rewritten against the code: parameter matrix, default precedence, tier correction and the qoder section; no runtime change) |
|
|
219
|
+
| [docs/release-v0.5.6.en.md](<docs/release-v0.5.6.en.md>) | v0.5.6 release notes (Qoder CN GUI adapter, hardware acceptance scope, macOS research boundary) |
|
|
220
|
+
| [docs/release-v0.5.5.en.md](<docs/release-v0.5.5.en.md>) | v0.5.5 release notes (Kimi Code GUI adapter: dual renderer processes, full-path workspace binding with native import, three-stage model selection and tier validation, run detection and recovery) |
|
|
217
221
|
| [docs/release-v0.5.4.en.md](<docs/release-v0.5.4.en.md>) | v0.5.4 release notes (optional AI visual content validation: user-supplied command delegation, majority-vote debouncing, warning-only by default) |
|
|
218
222
|
| [docs/release-v0.5.3.en.md](<docs/release-v0.5.3.en.md>) | v0.5.3 release notes (ZCode hardware-revisit fixes: instance survival across server exit, new-task page switch, send-failure attribution) |
|
|
219
223
|
| [docs/release-v0.5.2.en.md](<docs/release-v0.5.2.en.md>) | v0.5.2 release notes (ZCode project-less dispatch and `allowCreateProject`, issue #12) |
|
|
@@ -261,7 +265,7 @@ Use `server.log` when troubleshooting connections; do not treat stderr output it
|
|
|
261
265
|
- **Engineering / CI** ✅
|
|
262
266
|
- GitHub Actions: `CI` (`build-test` ubuntu/windows/macos × Node 20/22/24 + `pack-check`, plus a `visual-browser` real-browser matrix ubuntu/windows/macos-15-intel/macos-15 × Node 20/22/24, all green with the v0.5.1 tag) and `Release` (tag-triggered) both green
|
|
263
267
|
- Skill self-install verified idempotent on this machine's real `~/.rivet/skills/tianshu-mcp`
|
|
264
|
-
- npm package name `tianshu-mcp` published continuously since v0.1.1 (currently `0.5.
|
|
268
|
+
- npm package name `tianshu-mcp` published continuously since v0.1.1 (currently `0.5.7`)
|
|
265
269
|
- **Real Tianshu host integration (DoD #6)** ✅ (2026-09-07)
|
|
266
270
|
- Configured the local mode in the real `D:\Tianshu` desktop host `mcp.servers` → sidecar reported `MCP: 2 servers connected, 10 tools` (including this server's 8 tools), spawned the child process and connected over stdio
|
|
267
271
|
- Exposed and fixed a skill-install source-path bug (fileURLToPath, commit 55cf2d0)
|
|
@@ -352,6 +356,25 @@ Use `server.log` when troubleshooting connections; do not treat stderr output it
|
|
|
352
356
|
- **Fail-closed**: once enabled, an unresolvable command or a missing env reference yields a whole-round `configurationError` with no result rows; a single command failure produces only a blocked warning item that stays visible in the round message and in the repair plan's "warning-only items (no fix required)" section
|
|
353
357
|
- **Defect fix**: the repair plan no longer lists `optional:true` failures as "must fix"
|
|
354
358
|
- **CLI/diagnostics**: added `visual content probe <project> [ruleId]` and `visual content cache clear <taskId>`; `visual doctor` gained content-command resolution and budget-comparison findings
|
|
359
|
+
- **M24 — Kimi Code GUI adapter (fourth GUI agent) + v0.5.5** (2026-09-20) — **764 tests**
|
|
360
|
+
- **Dual-renderer CDP driver**: the model / thinking-tier / execution-mode menus render in the separate `Kimi Browser Overlay` window while the workspace menu and the "switch model" dialog stay in the main window
|
|
361
|
+
- **Full-path workspace binding**: an unregistered directory is imported through the native "add workspace" dialog (Win32 coordinate clicks plus `WM_SETTEXT`/`WM_GETTEXT` read-back), and same-name/different-directory cases fail closed
|
|
362
|
+
- **Three-stage model selection and tier validation**: pill read-back → overlay shortcut menu → "more models…" dialog; tiers are validated against **the set the UI actually renders** (official `Low/High/Max`, unofficial `On/Off`), and a tier the UI does not render fails before sending
|
|
363
|
+
- **Run detection and recovery**: `button.stop` / `send.is-starting` are the authoritative signals; the six `needs_user` kinds recover via `continue_task`, and an uncertain send is never repeated
|
|
364
|
+
- See the [Kimi Code guide](docs/kimi-cdp.en.md) and the [v0.5.5 release notes](<docs/release-v0.5.5.en.md>)
|
|
365
|
+
- **M25 — Qoder CN GUI adapter (fifth GUI agent) + v0.5.6** (2026-09-22) — **826 tests**
|
|
366
|
+
- **Installation discovery and instance reuse**: explicit `gui.exePath` → D-drive-first candidates → relative-path templates → standard directories, with Qoder CN identity checks; an existing instance without usable CDP is preserved in place and turned into `needs_user`, never closed or restarted
|
|
367
|
+
- **Workspaces and models**: full-path binding, with unregistered directories imported through New Task → Workspace → New Workspace → Add Read/Write Folder; `modelSource=default/custom` disambiguates cross-group name clashes; the thinking tier is saved in Model Management as a **global preference** and read back by reopening it, and an unsupported tier fails before sending
|
|
368
|
+
- **Send and answer checkpoints**: a checkpoint is written before sending the brief or submitting multi-question answers, and an unconfirmed receipt means observe-only, never an automatic resend; `continue_task` only reobserves for approval/login environment waits
|
|
369
|
+
- **Liveness and same-session repair**: completion must belong to this turn's user message (old replies, a static screen, and disconnects never qualify); both automatic and manual repair write the plan first, then send its filename, full path, and complete text to the original conversation
|
|
370
|
+
- **Platform boundary**: the Windows hardware loop passed (default/custom models, a newly registered workspace, and controlled failure → plan → same-session repair → re-acceptance); macOS is `research` and fail-closed
|
|
371
|
+
- See the [Qoder CN guide](docs/qoder-cdp.en.md) and the [v0.5.6 release notes](<docs/release-v0.5.6.en.md>)
|
|
372
|
+
- **M26 — Orchestration skill docs rewritten against the code + v0.5.7** (2026-09-22) — **826 tests** (no runtime change)
|
|
373
|
+
- **Parameter compatibility matrix**: `projectPath`/`model`/`modelSource`/`reasoningLevel`/`mode`/`planDoc`/`designSystem`/`allowCreateProject`/`continue_task` across the five built-in agents, stating plainly that a wrong parameter is rejected
|
|
374
|
+
- **Default precedence**: `autoVerify` (on by default at the server level), `autoFixRounds` (codex 5 / zcode 2 / kimicode 2 / qoder 3) and `taskTimeoutMs`
|
|
375
|
+
- **Kimi Code tier domain corrected** (`低/low`, `高/high`, `max`, `on`, `off` — deliberately without `中`/`medium`) and the qoder section completed (`modelSource` disambiguation, tier saved then read back, checkpoints that prevent resends, repair writing a plan before sending its full text, macOS dispatch disabled)
|
|
376
|
+
- **New `needsUserKind` × agent × `continue_task` matrix** and an `agentEndReason` → terminal-state mapping, plus `project_not_registered`/`unsupported_platform`/`qoder_error` and the new meta fields
|
|
377
|
+
- See the [v0.5.7 release notes](<docs/release-v0.5.7.en.md>)
|
|
355
378
|
|
|
356
379
|
## Agent support status
|
|
357
380
|
|
|
@@ -428,7 +451,7 @@ Behavior and limits:
|
|
|
428
451
|
|
|
429
452
|
| Document | Content |
|
|
430
453
|
|---|---|
|
|
431
|
-
| [CHANGELOG.en.md](<CHANGELOG.en.md>) | Version history (v0.1.0 → v0.5.
|
|
454
|
+
| [CHANGELOG.en.md](<CHANGELOG.en.md>) | Version history (v0.1.0 → v0.5.7) |
|
|
432
455
|
| [CONTRIBUTING.en.md](CONTRIBUTING.en.md) | Dev setup, conventions, commit/release flow, adding an agent |
|
|
433
456
|
| [SECURITY.en.md](SECURITY.en.md) | Security model (zero credentials / command whitelist / process & desktop-automation boundaries) and private reporting |
|
|
434
457
|
| [CODE_OF_CONDUCT.en.md](CODE_OF_CONDUCT.en.md) | Contributor Code of Conduct |
|
package/README.md
CHANGED
|
@@ -204,13 +204,17 @@ run_task(projectPath=D:/xxx/my-app, agentId=kimicode, task=「按 `./plan.md`
|
|
|
204
204
|
| [ARCHITECTURE.md](ARCHITECTURE.md) | **架构说明**:分层模型与模块边界、启动装配、数据目录、状态机、验收与返修流水线、Agent 驱动层契约、GUI 实例生命周期、跨平台策略、安全红线、扩展点、已知缺口 |
|
|
205
205
|
| [docs/tianshu-integration.md](docs/tianshu-integration.md) | 天枢 config.json 两种接入模式、UI/API 操作、冒烟步骤、FAQ |
|
|
206
206
|
| [docs/agent-profiles.md](docs/agent-profiles.md) | agent profiles 字段说明 + 真实机器样例(codex M2 定稿) |
|
|
207
|
-
| [docs/adapter-matrix.md](docs/adapter-matrix.md) | 各 Agent 能力调研矩阵(Codex/Zcode/TraeWork/Kimi Code/扩展位) |
|
|
207
|
+
| [docs/adapter-matrix.md](docs/adapter-matrix.md) | 各 Agent 能力调研矩阵(Codex/Zcode/TraeWork/Kimi Code/Qoder CN/扩展位) |
|
|
208
208
|
| [docs/traework-cdp.md](docs/traework-cdp.md) | TraeWork GUI 驱动(CDP):原理、配置、模式切换、选择器、安全红线、踩坑记录、验证记录 |
|
|
209
209
|
| [docs/zcode-cdp.md](docs/zcode-cdp.md) | ZCode GUI 驱动:安装探测、精确项目/模型、完全访问、暂停继续、验收返修与双平台状态 |
|
|
210
210
|
| [docs/zcode-windows-smoke.md](docs/zcode-windows-smoke.md) | ZCode Windows 真机开发、同会话返修与提问续跑验收记录 |
|
|
211
211
|
| [docs/codex-gui-cdp.md](docs/codex-gui-cdp.md) | Codex 桌面端 GUI 驱动:MSIX COM 激活、CDP 接管、选择器、运行检测、验收返修 |
|
|
212
212
|
| [docs/kimi-cdp.md](docs/kimi-cdp.md) | Kimi Code GUI 驱动:双渲染进程(主窗口 + `Kimi Browser Overlay`)、工作区完整路径绑定与原生对话框导入、模型三级选择与思考档位、执行模式、运行检测与排障 |
|
|
213
213
|
| [docs/codex-windows-smoke.md](docs/codex-windows-smoke.md) | Codex Windows 真机验收记录(含验收失败→自动生成计划→返修通过闭环) |
|
|
214
|
+
| [docs/qoder-cdp.md](docs/qoder-cdp.md) | Qoder CN GUI 驱动:安装发现与实例复用、完整路径工作区与原生导入、`modelSource` 与模型管理全局思考等级、发送/答题检查点、运行判定与原会话返修、真机证据与未覆盖项 |
|
|
215
|
+
| [docs/release-v0.5.7.md](<docs/release-v0.5.7.md>) | v0.5.7 发布说明(编排技能文档按代码实况重写:参数兼容矩阵、默认值优先级、档位修正与 qoder 章节;无运行时变更) |
|
|
216
|
+
| [docs/release-v0.5.6.md](<docs/release-v0.5.6.md>) | v0.5.6 发布说明(Qoder CN GUI 适配、真机验收范围与 macOS research 边界) |
|
|
217
|
+
| [docs/release-v0.5.5.md](<docs/release-v0.5.5.md>) | v0.5.5 发布说明(Kimi Code GUI 适配:双渲染进程、工作区完整路径绑定与原生导入、模型三级选择与档位校验、运行检测与恢复) |
|
|
214
218
|
| [docs/release-v0.5.4.md](<docs/release-v0.5.4.md>) | v0.5.4 发布说明(可选 AI 视觉内容校验:自备命令委托、多数票防抖、默认仅告警) |
|
|
215
219
|
| [docs/release-v0.5.3.md](<docs/release-v0.5.3.md>) | v0.5.3 发布说明(ZCode 真机回访修复:实例跨 server 驻留、新建任务切页、发送失败归因) |
|
|
216
220
|
| [docs/release-v0.5.2.md](<docs/release-v0.5.2.md>) | v0.5.2 发布说明(ZCode 无项目派发与 `allowCreateProject`,issue #12) |
|
|
@@ -262,7 +266,7 @@ run_task(projectPath=D:/xxx/my-app, agentId=kimicode, task=「按 `./plan.md`
|
|
|
262
266
|
- **工程 / CI** ✅
|
|
263
267
|
- GitHub Actions:`CI`(`build-test` ubuntu/windows/macos × Node 20/22/24 + `pack-check`,另加 `visual-browser` 真实浏览器矩阵 ubuntu/windows/macos-15-intel/macos-15 × Node 20/22/24,随 v0.5.1 tag 全绿)与 `Release`(tag 触发)均绿
|
|
264
268
|
- 技能自检安装已在本机真实 `~/.rivet/skills/tianshu-mcp` 验证生效且幂等
|
|
265
|
-
- npm 包名 `tianshu-mcp` 自 v0.1.1 起持续发布(当前 `0.5.
|
|
269
|
+
- npm 包名 `tianshu-mcp` 自 v0.1.1 起持续发布(当前 `0.5.7`)
|
|
266
270
|
- **天枢宿主真实接入(DoD #6)** ✅(2026-09-07,[host-integration-record.md](docs/host-integration-record.md))
|
|
267
271
|
- 在真实 `D:\Tianshu` 桌面宿主 `mcp.servers` 配置本地模式 → sidecar `MCP: 2 servers connected, 10 tools`(含本 server 8 工具),spawn 子进程并 stdio 连通
|
|
268
272
|
- 实测暴露并修复技能安装源路径 bug(fileURLToPath,提交 55cf2d0)
|
|
@@ -353,6 +357,25 @@ run_task(projectPath=D:/xxx/my-app, agentId=kimicode, task=「按 `./plan.md`
|
|
|
353
357
|
- **fail-closed**:启用后命令不可解析/env 引用缺失 → 整轮 `configurationError` 且不产出结果行;单项命令失败仅产生 blocked 告警项,并在整轮消息与返修计划的「仅告警项(不必修复)」小节可见
|
|
354
358
|
- **缺陷修复**:返修计划不再把 `optional:true` 的失败列为「必须修复」
|
|
355
359
|
- **CLI/诊断**:新增 `visual content probe <project> [ruleId]` 与 `visual content cache clear <taskId>`;`visual doctor` 新增内容命令解析与预算对比两项 finding
|
|
360
|
+
- **M24 — Kimi Code GUI 适配(第四个 GUI agent)+ v0.5.5**(2026-09-20)— **764 测试**
|
|
361
|
+
- **双渲染进程 CDP 驱动**:模型 / 思考档位 / 执行模式菜单渲染在独立的 `Kimi Browser Overlay` 浮层窗口,工作区菜单与「切换模型」对话框仍在主窗口
|
|
362
|
+
- **工作区完整路径绑定**:未登记目录经原生「添加工作区」对话框导入(Win32 坐标点击 + `WM_SETTEXT`/`WM_GETTEXT` 回读),同名不同目录一律 fail-closed
|
|
363
|
+
- **模型三级选择与档位校验**:pill 回读 → overlay 快捷菜单 → 「更多模型…」对话框;档位按**界面实际渲染的集合**校验(官方 `Low/High/Max`,非官方 `On/Off`),请求不存在的档位在发送前报错
|
|
364
|
+
- **运行检测与恢复**:`button.stop` / `send.is-starting` 为权威信号;`needs_user` 六类由 `continue_task` 恢复,发布发送确认失败绝不重发
|
|
365
|
+
- 详见 [Kimi Code 文档](docs/kimi-cdp.md) 与 [v0.5.5 发布说明](<docs/release-v0.5.5.md>)
|
|
366
|
+
- **M25 — Qoder CN GUI 适配(第五个 GUI agent)+ v0.5.6**(2026-09-22)— **826 测试**
|
|
367
|
+
- **安装发现与实例复用**:显式 `gui.exePath` → D 盘优先候选 → 相对路径模板 → 标准目录;校验 Qoder CN 身份;已有实例无可用 CDP 时保留现场转 `needs_user`,绝不关闭或重启
|
|
368
|
+
- **工作区与模型**:完整路径绑定,未登记目录经「新的任务 → 工作区 → 新建工作区 → 添加可读写文件夹」原生导入;`modelSource=default/custom` 消除跨组重名;思考等级经「模型管理」保存为**全局偏好**并重新打开回读,不支持的档位在发送前报错
|
|
369
|
+
- **发送与答题检查点**:发送任务书与提交多题答案前落检查点,未确认回执时只观察、不自动重发;`continue_task` 对审批/登录等环境等待只恢复观察
|
|
370
|
+
- **运行判定与原会话返修**:完成必须绑定本轮用户消息(历史回复、界面静止、断线均不算);自动与手动返修都先落修复计划,再把文件名、完整路径与全文发送原会话
|
|
371
|
+
- **平台边界**:Windows 真机闭环通过(默认/自定义模型、新登记工作区、受控失败 → 落计划 → 原会话返修 → 再验收);macOS 为 `research` 且 fail-closed
|
|
372
|
+
- 详见 [Qoder CN 文档](docs/qoder-cdp.md) 与 [v0.5.6 发布说明](<docs/release-v0.5.6.md>)
|
|
373
|
+
- **M26 — 编排技能文档按代码实况重写 + v0.5.7**(2026-09-22)— **826 测试**(无运行时变更)
|
|
374
|
+
- **参数兼容矩阵**:`projectPath`/`model`/`modelSource`/`reasoningLevel`/`mode`/`planDoc`/`designSystem`/`allowCreateProject`/`continue_task` 九个维度 × 五个内置 agent,写清「传错即报错」
|
|
375
|
+
- **默认值优先级**:`autoVerify`(server 默认开启)、`autoFixRounds`(codex 5 / zcode 2 / kimicode 2 / qoder 3)、`taskTimeoutMs` 取值顺序
|
|
376
|
+
- **修正 Kimi Code 档位取值域**(`低/low`、`高/high`、`max`、`on`、`off`,刻意不含 `中`/`medium`)与 qoder 章节(`modelSource` 消歧、档位保存回读、检查点不重发、返修先落计划再回发全文、macOS 禁止派发)
|
|
377
|
+
- **新增 `needsUserKind` × agent × `continue_task` 行为矩阵**与 `agentEndReason` → 终态映射,补 `project_not_registered`/`unsupported_platform`/`qoder_error` 等错误码与 meta 新字段
|
|
378
|
+
- 详见 [v0.5.7 发布说明](<docs/release-v0.5.7.md>)
|
|
356
379
|
|
|
357
380
|
## Agent 适配现状
|
|
358
381
|
|
|
@@ -430,7 +453,7 @@ run_task(projectPath=/path/to/项目, agentId=codex-cli, task="任务书", autoV
|
|
|
430
453
|
| 文档 | 内容 |
|
|
431
454
|
|---|---|
|
|
432
455
|
| [HANDOFF.md](HANDOFF.md) | 项目交接文档:当前状态快照、架构导览、硬性红线、已知限制、接手建议 |
|
|
433
|
-
| [CHANGELOG.md](<CHANGELOG.md>) | 版本变更日志(v0.1.0 → v0.5.
|
|
456
|
+
| [CHANGELOG.md](<CHANGELOG.md>) | 版本变更日志(v0.1.0 → v0.5.7) |
|
|
434
457
|
| [CONTRIBUTING.md](CONTRIBUTING.md) | 开发环境、工程规范、提交与发布流程、如何新增 agent |
|
|
435
458
|
| [SECURITY.md](SECURITY.md) | 安全模型(凭证零管理/命令白名单/进程与桌面自动化边界)与私密报告渠道 |
|
|
436
459
|
| [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md) | 贡献者行为准则 |
|