tianshu-mcp 0.3.2 → 0.3.4

package/CHANGELOG.en.md

@@ -19,9 +19,40 @@ Chinese version: [CHANGELOG.md](CHANGELOG.md)
 - Best-effort stop of a GUI-side pending session (via a temporary CDP connection) when cancelling
   a task in the `needs_user` state.
 
----
-
-## [0.3.2] — 2026-09-12
+---
+
+## [0.3.4] — 2026-09-13
+
+- Fix #8/#10 project-selector ambiguity, full-path binding false negatives and contaminated model readback.
+- Fix #9 environment recovery without a session: send full task/context/references once and locate the session through its marker or unique delta.
+- Add initialization recovery with shared deadlines, configurable budgets and a resumable `setup_recovery` state. Reconcile native side effects after timeouts and terminate helper processes on cancellation.
+- Handle split model labels, hidden outgoing animation text, residual project menus trapping focus and send buttons that are not yet ready.
+- Add real DOM and recovery/cancellation regressions. Windows hardware verifies cold first import, imported-project reuse and same-task recovery with actual artifacts and 2/2 acceptance checks.
+- Release/tarball only; no npm publication. macOS has no real-device verification for this patch and the ZCode profile remains `research`.
+
+See [release notes](docs/release-v0.3.4.en.md) and [validation evidence](docs/zcode-issue-8-10-validation.en.md).
+
+## [0.3.3] — 2026-09-12
+
+Fixes issues #4 / #7 by adapting the ZCode GUI driver to the 3.11.2 model-menu and project-binding semantics, while making acceptance fail closed for zero-test and zero-change outcomes.
+
+### Fixed
+
+- ZCode provider selectors now support both `group-provider` and the 3.11.2 `group-family` prefix. Model selection tries the visible model first and only expands a provider/family group as a fallback; failures include visible text and `data-testid` candidates.
+- New-project import dismisses the stale workspace menu before opening Add Project and uses a bounded three-round dismiss/click/verify loop, preventing the first click from being consumed as an outside click.
+- Project binding prioritizes the composer menu's `menuitemcheckbox`; the legacy sidebar item remains a compatibility fallback. Display-name matching normalizes NFKC, whitespace, and case.
+- Binding read-back combines the composer trigger text with the full-path fast path, rejects known unbound placeholders, and retries the idempotent bind operation for up to two rounds.
+- Built-in ZCode and TraeWork discovery paths use `{PROGRAMFILES}` / `{PROGRAMFILES(X86)}`. Environment placeholders in custom profiles are expanded case-insensitively while unknown placeholders remain unchanged.
+- A mandatory test check is failed when it exits with code 0 but its output reports zero executed tests.
+- Git projects now require a change relative to the pre-work baseline by default. Pure question/analysis tasks can explicitly opt out with `"requireChanges": false` in `.tianshu-mcp/acceptance.json`.
+
+### Tests
+
+- Regression coverage now models ZCode 3.11.2 flat family layouts, legacy provider fallback, testid diagnostics, swallowed project clicks, composer binding and read-back retries, plus zero-test/zero-change acceptance gates.
+
+---
+
+## [0.3.2] — 2026-09-12
 
 Fixes for issues #5 / #6: **the MCP task model was disconnected from the state of the turn inside
 the Codex GUI**. The former is the completion-detection deadlock that kept reporting `running`
@@ -494,7 +525,8 @@ project → pick model and reasoning level → send instructions → run detecti
 
 ---
 
-[Unreleased]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.2...HEAD
+[Unreleased]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.3...HEAD
+[0.3.3]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.2...v0.3.3
 [0.3.2]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.1...v0.3.2
 [0.3.1]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.0...v0.3.1
 [0.3.0]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.2.0...v0.3.0

package/CHANGELOG.md

@@ -17,9 +17,40 @@
 - Codex GUI 的 macOS 真机验证（当前内置状态为 `research`）。
 - needs_user 状态下取消任务时经临时 CDP 连接尽力停止 GUI 内等待中的会话。
 
----
-
-## [0.3.2] — 2026-09-12
+---
+
+## [0.3.4] — 2026-09-13
+
+- 修复 #8/#10 的项目选择器歧义、完整路径绑定误判和模型混合文本回读。
+- 修复 #9 的无会话环境恢复：完整任务、上下文和引用只发送一次，使用任务标记或唯一会话差集定位。
+- 增加共用截止时间的初始化恢复、可配置预算及 `setup_recovery` 暂停；原生操作超时先复检副作用，取消回收辅助进程。
+- 修复拆分模型标签、隐藏动画旧值、项目菜单焦点截获和发送按钮尚未就绪的问题。
+- 补充真实 DOM、取消及恢复回归，Windows 真机覆盖新项目冷导入、已导入项目复用和同任务恢复，实际产物与 2/2 验收通过。
+- 本版本发布 Release/tarball，不发布 npm；macOS 尚无本次真机验证，ZCode profile 保持 `research`。
+
+详见 [发布说明](docs/release-v0.3.4.md) 和 [验收证据](docs/zcode-issue-8-10-validation.md)。
+
+## [0.3.3] — 2026-09-12
+
+修复 issue #4 / #7：适配 ZCode 3.11.2 的模型菜单与项目绑定语义，并将验收引擎收紧为 fail-closed，防止零用例、零变更任务假绿。
+
+### 修复
+
+- ZCode 供应商分组选择器同时兼容 `group-provider` 与 3.11.2 `group-family`；模型流程改为直选优先、供应商分组展开兜底，并在失败信息中附带可见文本与 `data-testid` 候选。
+- 添加新项目前收起残留工作区菜单，以最多三轮“收起—点击—验证”闭环消除首次点击被吞。
+- 项目选择优先点击 composer 菜单的 `menuitemcheckbox`，旧版侧栏项仅作回退；项目名称按 NFKC、空白和大小写归一化。
+- 项目绑定回读同时校验 composer 触发器文本与完整路径，未绑定占位词不再误判；绑定失败最多两轮幂等重试。
+- ZCode/TraeWork 内置发现目录统一为 `{PROGRAMFILES}` / `{PROGRAMFILES(X86)}`；自定义 profile 的环境占位符按大小写不敏感方式展开，未知占位符保留原样。
+- 非 optional 测试检查在退出码为 0 但输出显示零用例时改判失败。
+- Git 项目默认要求相对动工前基线产生变更；纯问答/分析任务可在 `.tianshu-mcp/acceptance.json` 设置 `"requireChanges": false` 显式关闭。
+
+### 测试
+
+- 回归覆盖 ZCode 3.11.2 family 平铺模型、旧版分组兜底、testid 诊断、项目点击被吞、composer 绑定与回读重试，以及零用例/零变更门禁。
+
+---
+
+## [0.3.2] — 2026-09-12
 
 修复 issue #5 / #6：**MCP 任务模型与 Codex GUI 内 turn 的状态脱节**。
 前者是"等用户确认时恒报 running"的完成判定死锁，后者是 `cancel_task` 只停 MCP 侧等待、
@@ -431,7 +462,8 @@ Codex 桌面端改为 **GUI 驱动**：新增 `codex-gui` adapter，通过 MSIX
 
 ---
 
-[未发布]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.2...HEAD
+[未发布]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.3...HEAD
+[0.3.3]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.2...v0.3.3
 [0.3.2]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.1...v0.3.2
 [0.3.1]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.3.0...v0.3.1
 [0.3.0]: https://github.com/lanlan0811/tianshu-mcp/compare/v0.2.0...v0.3.0

package/README.en.md

@@ -36,7 +36,7 @@ Tianshu plays the role of the overall commander; this MCP server is the **schedu
 
 - **9 MCP tools**: `run_task / continue_task / query_task / list_tasks / get_task_report / cancel_task / verify_task / rework_task / get_profiles`.
 - **Async contract**: `run_task` returns a `taskId` immediately; long-running work is polled via `query_task` (never blocks `tools/call`).
-- **Objective acceptance**: automated command checks (typecheck/lint/test/build — skipped when absent, plus tech-stack derivation) + programmatic code analysis (changed-file list / diffstat / suspicious signals such as TODO, debugger, secret-like patterns), all relative to a **git baseline**; never auto-commits or stashes.
+- **Objective acceptance**: automated command checks (typecheck/lint/test/build — skipped when absent, plus tech-stack derivation) + programmatic code analysis (changed-file list / diffstat / suspicious signals such as TODO, debugger, secret-like patterns), all relative to a **git baseline**; never auto-commits or stashes. The acceptance engine is **fail-closed**: a test check fails when its output reports zero executed tests even if the exit code is 0; git projects must produce changes relative to the pre-work baseline by default (pure analysis tasks can opt out with `"requireChanges": false` in `.tianshu-mcp/acceptance.json`).
 - **Rework loop**: automatic rework (`autoFixRounds`) + manual `rework_task`; on verification failure a repair-plan file is generated and fed back to the agent; when rounds run out → `needs_attention` awaiting Tianshu's verdict.
 - **Execution surfaces**: `driver: "gui"` selects an explicit, isolated Codex/TraeWork/ZCode CDP adapter; `driver: "spawn"` runs an external CLI child process.
 - **Scheduling discipline**: per-project serial queue + global concurrency cap (default 2, configurable).
@@ -63,7 +63,7 @@ git clone https://github.com/lanlan0811/tianshu-mcp.git
 cd tianshu-mcp
 npm ci
 npm run build        # sync-version + tsc → dist/
-npm test             # 340 tests across 39 files, including Codex/ZCode unit/fake-CDP/restart/repair coverage
+npm test             # 366 tests across 39 files, including Codex/ZCode unit/fake-CDP/restart/repair coverage
 ```
 
 ### Install the npm package
@@ -122,6 +122,11 @@ run_task(projectPath=D:/xxx/my-app, task=「…task brief…」, agentId=codex,
 > COM activation with a dedicated `--user-data-dir` before CDP can drive it. `planDoc` / `designSystem` are
 > appended to the initial instruction. See [docs/codex-gui-cdp.en.md](docs/codex-gui-cdp.en.md) and the
 > [hardware acceptance record](docs/codex-windows-smoke.en.md). Projects not yet registered on the Codex side are **registered automatically** — no manual project creation needed.
+> When Codex parks on a "waiting for user confirmation" screen (plan-confirmation card, subscription checkout, etc.),
+> the task turns `needs_user(user_confirmation)`; after handling it in the Codex window call `continue_task(taskId)`
+> to resume observation. `cancel_task` clicks the in-app stop control over CDP and bounded-waits for the GUI to go
+> idle; before dispatching, a still-running managed instance is stopped best-effort, and if it never goes idle the
+> dispatch is rejected with `instance_busy`.
 
 When driving TraeWork, `model` and `mode` are available:
 
@@ -140,18 +145,18 @@ run_task(projectPath=D:/xxx/my-app, agentId=zcode, task="Implement `./plan.md`",
          model=DeepSeek/deepseek-flash, autoVerify=true)
 ```
 
-Questions, login, an existing non-CDP instance, or system permission pause as `needs_user`; call `continue_task(taskId, message)` to resume the recorded session. See [docs/zcode-cdp.en.md](docs/zcode-cdp.en.md).
+Questions, login, an existing non-CDP instance, or system permission pause as `needs_user`; call `continue_task(taskId, message)` to resume the recorded session. Model selection is adapted to ZCode 3.11.2: flat models are selected directly first, with provider/family group expansion as a fallback — both new and legacy layouts are supported. See [docs/zcode-cdp.en.md](docs/zcode-cdp.en.md).
 
 ## Tool surface (9 tools)
 
 | Tool | Capability / approval | Purpose |
 |---|---|---|
 | `run_task` | write + approval | Dispatch work (optional auto-verify / auto-rework); returns `taskId` asynchronously |
-| `continue_task` | write + approval | Resume the original ZCode session from `needs_user` |
+| `continue_task` | write + approval | Resume the session behind `needs_user` (ZCode resumes the recorded session; Codex re-observes for `user_confirmation` / re-dispatches for `login_required`) |
 | `query_task` | read | Poll status / progress / log tail |
 | `list_tasks` | read | Filtered history of tasks |
 | `get_task_report` | read | Full text of a verification round's report (`report.md`) |
-| `cancel_task` | write + approval | Cancel a running task (kill process tree) |
+| `cancel_task` | write + approval | Cancel a running task: CLI agents kill the process tree; GUI agents click the in-app stop control over CDP and bounded-wait (`gui.cancelWaitMs`, default 15s) for the GUI to go idle, stating so explicitly in the final message when the stop is unconfirmed |
 | `verify_task` | read | Run one verification pass on a task/project path (no source changes) |
 | `rework_task` | write + approval | Manual rework (feed the failure report back to the same agent) |
 | `get_profiles` | read | Inspect agent adapters and executable discovery results |
@@ -182,6 +187,8 @@ Use `server.log` when troubleshooting connections; do not treat stderr output it
 | [docs/zcode-windows-smoke.en.md](docs/zcode-windows-smoke.en.md) | ZCode Windows hardware record for development, same-session repair, and question continuation |
 | [docs/codex-gui-cdp.en.md](docs/codex-gui-cdp.en.md) | Codex desktop GUI driver: MSIX COM activation, CDP attach, selectors, run detection, verify/repair |
 | [docs/codex-windows-smoke.en.md](docs/codex-windows-smoke.en.md) | Codex Windows hardware record (incl. verify-fail → auto plan → repair-pass loop) |
+| [docs/release-v0.3.3.en.md](docs/release-v0.3.3.en.md) | v0.3.3 release notes (ZCode 3.11.2 adaptation + fail-closed acceptance engine) |
+| [docs/release-v0.3.2.en.md](docs/release-v0.3.2.en.md) | v0.3.2 release notes (Codex wait-user detection + cancel truly stops the GUI) |
 | [docs/release-v0.3.1.en.md](docs/release-v0.3.1.en.md) | v0.3.1 release notes (skill docs rewrite + release automation fixes) |
 | [docs/release-v0.3.0.en.md](docs/release-v0.3.0.en.md) | v0.3.0 release notes (Codex desktop GUI adapter, incl. BREAKING) |
 | [docs/release-v0.2.0.en.md](docs/release-v0.2.0.en.md) | v0.2.0 release notes (unified ZCode GUI loop) |
@@ -258,14 +265,21 @@ Use `server.log` when troubleshooting connections; do not treat stderr output it
 - **M13 — Skill docs alignment + release automation fixes + v0.3.1** (2026-09-12, see [release-v0.3.1.en.md](docs/release-v0.3.1.en.md))
   - Self-installed skill docs (SKILL.md / usage-examples.md) rewritten item by item against the v0.3.0 tool surface (codex GUI params, needs_user handling, verify/list usage)
   - Bilingual release bodies, Full Changelog & CI link fixes, and automated Gitee releases
+- **M14 — Codex wait-user detection + cancel truly stops the GUI + v0.3.2** (2026-09-12, fixes issues #5 / #6, see [release-v0.3.2.en.md](docs/release-v0.3.2.en.md))
+  - Issue #5: Codex parked on a "waiting for user confirmation" screen no longer dead-locks in `running` — with the stop button visible and the conversation hash unchanged for `gui.stallTimeoutMs` (default 5 min), the task turns `needs_user(user_confirmation)`; added a configurable `gui.selectors.userGate` UI gate; `continue_task` now supports codex (`user_confirmation` re-observes / `login_required` re-dispatches)
+  - Issue #6: `cancel_task` for GUI agents clicks the in-app stop control over CDP and bounded-waits (`gui.cancelWaitMs`, default 15s) for the GUI to go idle before recording `cancelled`; dispatch-time liveness checks reject with `instance_busy` when a managed instance is still running, preventing overlapping turns
+- **M15 — ZCode 3.11.2 adaptation + fail-closed acceptance engine + v0.3.3** (2026-09-12, fixes issues #4 / #7, see [release-v0.3.3.en.md](docs/release-v0.3.3.en.md)) — **366 tests**
+  - Issue #4: the model menu supports both `group-provider` and 3.11.2 `group-family` groups, selecting flat models directly first with group expansion as a fallback; project binding now keys on the composer's `menuitemcheckbox` with read-back checking trigger text + full path and up to two idempotent retries; stale workspace menus are dismissed before adding a project
+  - Issue #7: a test check that exits 0 but reports zero executed tests fails; git projects must produce changes relative to the pre-work baseline by default (opt out with `requireChanges: false`) — zero tests and zero changes no longer pass silently
+  - `{PROGRAMFILES}` placeholders normalized to uppercase and environment-variable expansion made case-insensitive
 
 ## Agent support status
 
-| agentId | driver | status | Notes |
+| agentId | driver / adapter | status | Notes |
 |---|---|---|---|
-| `codex` | `gui` | **ready** | Desktop GUI over CDP (MSIX COM activation + dedicated profile); supports `model`/`reasoningLevel`/`planDoc`/`designSystem`; Windows machine-verified |
-| `zcode` | `zcode-gui` | **research** | CDP GUI adapter and Windows hardware loop passed; remains non-ready until macOS hardware passes |
-| `traework` | **`gui`** | **ready** | CDP-driven TRAE SOLO CN desktop UI; all three panel modes machine-verified |
+| `codex` | `gui` / `codex-gui` | **ready** | Desktop GUI over CDP (MSIX COM activation + dedicated profile); supports `model`/`reasoningLevel`/`planDoc`/`designSystem`; wait-user, cancel and dispatch-guard semantics machine-verified (v0.3.2); Windows machine-verified |
+| `zcode` | `gui` / `zcode-gui` | **research** | CDP GUI adapter with the Windows hardware loop passed; adapted to ZCode 3.11.2 model menu and project binding (v0.3.3); remains non-ready until macOS hardware passes |
+| `traework` | `gui` / `traework-gui` | **ready** | CDP-driven TRAE SOLO CN desktop UI; all three panel modes machine-verified |
 | `stub` | `spawn` | tests only | `test/stub-agent/stub-agent.mjs` with 3 playbooks (good/fix-on-first/never) |
 
 > Adding an agent usually needs only a profile — see [docs/agent-profiles.en.md](docs/agent-profiles.en.md) and [CONTRIBUTING.en.md](CONTRIBUTING.en.md).
@@ -280,7 +294,7 @@ Use `server.log` when troubleshooting connections; do not treat stderr output it
 
 | Document | Content |
 |---|---|
-| [CHANGELOG.en.md](CHANGELOG.en.md) | Version history (v0.1.0 → v0.3.1) |
+| [CHANGELOG.en.md](CHANGELOG.en.md) | Version history (v0.1.0 → v0.3.3) |
 | [CONTRIBUTING.en.md](CONTRIBUTING.en.md) | Dev setup, conventions, commit/release flow, adding an agent |
 | [SECURITY.en.md](SECURITY.en.md) | Security model (zero credentials / command whitelist / process & desktop-automation boundaries) and private reporting |
 | [CODE_OF_CONDUCT.en.md](CODE_OF_CONDUCT.en.md) | Contributor Code of Conduct |

package/README.md

@@ -36,7 +36,7 @@
 
 - **9 个 MCP 工具**：`run_task / continue_task / query_task / list_tasks / get_task_report / cancel_task / verify_task / rework_task / get_profiles`。
 - **异步契约**：`run_task` 秒回 `taskId`，长任务用 `query_task` 轮询（长任务不卡 `tools/call`）。
-- **客观验收**：自动命令检查（typecheck/lint/test/build，缺则跳过 + 技术栈推导）+ 程序化代码分析（变更清单/diffstat/TODO·debugger·密钥形态等可疑标记），全部相对 **git 基线**，不自动 commit/stash。
+- **客观验收**：自动命令检查（typecheck/lint/test/build，缺则跳过 + 技术栈推导）+ 程序化代码分析（变更清单/diffstat/TODO·debugger·密钥形态等可疑标记），全部相对 **git 基线**，不自动 commit/stash。验收引擎 **fail-closed**：测试命令退出码为 0 但输出显示零用例时判失败；git 项目默认要求相对动工前基线产生变更（纯分析任务可在 `.tianshu-mcp/acceptance.json` 设 `"requireChanges": false` 显式关闭）。
 - **失败返修闭环**：自动返修（`autoFixRounds`）+ 手动 `rework_task`；验收失败时自动生成修复计划文件并回填给 agent；轮次用尽 → `needs_attention` 等天枢裁决。
 - **执行面**：`driver: "gui"` 由显式 adapter 驱动桌面 UI（Codex / TraeWork / ZCode 各自使用隔离的 CDP 流程）；`driver: "spawn"` 走外部 CLI 子进程。
 - **调度纪律**：每项目串行队列 + 全局并发上限（默认 2，可配）。
@@ -63,7 +63,7 @@ git clone https://github.com/lanlan0811/tianshu-mcp.git
 cd tianshu-mcp
 npm ci
 npm run build        # sync-version + tsc → dist/
-npm test             # 340 项测试：39 个文件，含 Codex/ZCode 单元/假 CDP/重启/返修闭环
+npm test             # 366 项测试：39 个文件，含 Codex/ZCode 单元/假 CDP/重启/返修闭环
 ```
 
 ### 安装 npm 包
@@ -120,6 +120,9 @@ run_task(projectPath=D:/xxx/my-app, task=「…任务书…」, agentId=codex,
 > 其 `ChatGPT.exe` 无法直接启动（被策略拒绝），须经 COM 激活并注入专属 `--user-data-dir` 后方可
 > 用 CDP 驱动。可传 `planDoc` / `designSystem` 拼进初始指令。详见 [docs/codex-gui-cdp.md](docs/codex-gui-cdp.md)
 > 与 [真机验收记录](docs/codex-windows-smoke.md)。项目未在 Codex 侧登记时会**自动登记**，无需手动建项目。
+> Codex 停在「等待用户确认」界面（方案确认卡/订阅结账页等）会转 `needs_user(user_confirmation)`，
+> 用户在 Codex 窗口处理完后调 `continue_task(taskId)` 恢复观察；`cancel_task` 会经 CDP 点击停止并
+> 有界等待 GUI 空闲，派发前若受管实例仍在运行会先尽力停止，仍不空闲则以 `instance_busy` 拒绝派发。
 
 驱动 TraeWork 时可用 `model` 与 `mode`：
 
@@ -138,18 +141,18 @@ run_task(projectPath=D:/xxx/my-app, agentId=zcode, task=「按 `./plan.md` 完
          model=DeepSeek/deepseek-flash, autoVerify=true)
 ```
 
-ZCode 提问或需要用户处理登录、旧实例、系统权限时进入 `needs_user`；处理后调用 `continue_task(taskId, message)` 恢复原会话。完整约束见 [docs/zcode-cdp.md](docs/zcode-cdp.md)。
+ZCode 提问或需要用户处理登录、旧实例、系统权限时进入 `needs_user`；处理后调用 `continue_task(taskId, message)` 恢复原会话。模型选择已适配 ZCode 3.11.2：直选平铺模型优先，展开 provider/family 分组兜底，新旧布局均兼容。完整约束见 [docs/zcode-cdp.md](docs/zcode-cdp.md)。
 
 ## 工具面（9 个）
 
 | 工具 | 能力 / 审批 | 作用 |
 |---|---|---|
 | `run_task` | write + 审批 | 派活（可带自动验收/自动返修），异步返回 `taskId` |
-| `continue_task` | write + 审批 | 恢复 `needs_user` 的原 ZCode 会话 |
+| `continue_task` | write + 审批 | 恢复 `needs_user` 的原会话（ZCode 恢复原会话；Codex 按 `user_confirmation` 重新观察 / `login_required` 重派） |
 | `query_task` | read | 轮询状态 / 进度 / 日志尾 |
 | `list_tasks` | read | 历史任务过滤列表 |
 | `get_task_report` | read | 某轮验收报告全文（`report.md`） |
-| `cancel_task` | write + 审批 | 取消运行中任务（kill 进程树） |
+| `cancel_task` | write + 审批 | 取消运行中任务：CLI agent kill 进程树；GUI agent 经 CDP 点击停止并在 `gui.cancelWaitMs`（默认 15s）内有界等待 GUI 空闲，未确认停止时终态明示 |
 | `verify_task` | read | 对任务/项目路径做一次验收（不改源码） |
 | `rework_task` | write + 审批 | 手动返修（把失败报告喂回同一 agent） |
 | `get_profiles` | read | 查看 agent 适配与可执行探测结果 |
@@ -180,6 +183,8 @@ ZCode 提问或需要用户处理登录、旧实例、系统权限时进入 `nee
 | [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 |
 | [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 |
 | [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） |
+| [docs/release-v0.3.3.md](docs/release-v0.3.3.md) | v0.3.3 发布说明（ZCode 3.11.2 适配 + 验收引擎 fail-closed） |
+| [docs/release-v0.3.2.md](docs/release-v0.3.2.md) | v0.3.2 发布说明（Codex 等待用户检测 + cancel 真停 GUI） |
 | [docs/release-v0.3.1.md](docs/release-v0.3.1.md) | v0.3.1 发布说明（技能文档重写 + 发布自动化修复） |
 | [docs/release-v0.3.0.md](docs/release-v0.3.0.md) | v0.3.0 发布说明（Codex 桌面端 GUI 适配，含 BREAKING） |
 | [docs/release-v0.2.0.md](docs/release-v0.2.0.md) | v0.2.0 发布说明（ZCode GUI 统一闭环） |
@@ -258,14 +263,21 @@ ZCode 提问或需要用户处理登录、旧实例、系统权限时进入 `nee
 - **M13 — 技能文档对齐 + 发布自动化修复 + v0.3.1**（2026-09-12，见 [release-v0.3.1.md](docs/release-v0.3.1.md)）
   - 技能自检安装文档（SKILL.md / usage-examples.md）对照 v0.3.0 工具面逐项重写（codex GUI 参数、needs_user 处理、verify/list 用法）
   - Release 正文双语合成、Full Changelog 与 CI 链接修复、Gitee 发行版纳入自动化
+- **M14 — Codex 等待用户检测 + 取消真停 GUI + v0.3.2**（2026-09-12，修复 issue #5 / #6，见 [release-v0.3.2.md](docs/release-v0.3.2.md)）
+  - issue #5：Codex 停在「等待用户确认」界面不再死锁在 `running`——停止按钮可见且对话哈希 `gui.stallTimeoutMs`（默认 5 分钟）不变 → 转 `needs_user(user_confirmation)`；新增可配置 `gui.selectors.userGate` 界面检测；`continue_task` 扩展支持 codex（`user_confirmation` 重新观察 / `login_required` 重派）
+  - issue #6：`cancel_task` 对 GUI agent 经 CDP 尽力点击停止并在 `gui.cancelWaitMs`（默认 15s）内有界等待 GUI 空闲后才落 `cancelled`；派发前检测受管实例运行态，仍运行则以 `instance_busy` 拒绝，杜绝新旧 turn 交叠
+- **M15 — ZCode 3.11.2 适配 + 验收引擎 fail-closed + v0.3.3**（2026-09-12，修复 issue #4 / #7，见 [release-v0.3.3.md](docs/release-v0.3.3.md)）— **366 测试**
+  - issue #4：模型菜单同时兼容 `group-provider` 与 3.11.2 `group-family` 分组，直选平铺模型优先、分组展开兜底；项目绑定改以 composer 复选项为主判据，回读校验触发器文本 + 完整路径，失败最多两轮幂等重试；添加项目前先收起残留菜单并重试
+  - issue #7：测试检查退出码 0 但输出零用例时改判失败；git 项目默认要求相对基线产生变更（`requireChanges: false` 可显式关闭），零用例与零变更不再假绿
+  - `{PROGRAMFILES}` 占位符统一大写且环境变量展开大小写不敏感
 
 ## Agent 适配现状
 
-| agentId | driver | status | 说明 |
+| agentId | driver / adapter | status | 说明 |
 |---|---|---|---|
-| `codex` | **`gui`** | **ready** | Codex 桌面端 GUI（MSIX COM 激活 + CDP）；支持 `model`/`reasoningLevel`/`planDoc`/`designSystem`；Windows 真机已验证 |
-| `zcode` | `zcode-gui` | **research** | CDP GUI adapter 已实现且 Windows 真机闭环通过；macOS 真机完成前不标 `ready` |
-| `traework` | **`gui`** | **ready** | CDP 驱动 TRAE SOLO CN 桌面 UI；三种面板模式真机验证通过 |
+| `codex` | `gui` / `codex-gui` | **ready** | Codex 桌面端 GUI（MSIX COM 激活 + CDP）；支持 `model`/`reasoningLevel`/`planDoc`/`designSystem`；等待用户确认、取消与重派护栏均已真机验证（v0.3.2）；Windows 真机已验证 |
+| `zcode` | `gui` / `zcode-gui` | **research** | CDP GUI adapter 已实现且 Windows 真机闭环通过；已适配 ZCode 3.11.2 模型菜单与项目绑定（v0.3.3）；macOS 真机完成前不标 `ready` |
+| `traework` | `gui` / `traework-gui` | **ready** | CDP 驱动 TRAE SOLO CN 桌面 UI；三种面板模式真机验证通过 |
 | `stub` | `spawn` | 仅测试 | `test/stub-agent/stub-agent.mjs` 三剧本（good/fix-on-first/never） |
 
 > 新增 agent 通常只需加一个 profile，详见 [docs/agent-profiles.md](docs/agent-profiles.md) 与 [CONTRIBUTING.md](CONTRIBUTING.md)。
@@ -281,7 +293,7 @@ ZCode 提问或需要用户处理登录、旧实例、系统权限时进入 `nee
 | 文档 | 内容 |
 |---|---|
 | [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 |
-| [CHANGELOG.md](CHANGELOG.md) | 版本变更日志（v0.1.0 → v0.3.1） |
+| [CHANGELOG.md](CHANGELOG.md) | 版本变更日志（v0.1.0 → v0.3.3） |
 | [CONTRIBUTING.md](CONTRIBUTING.md) | 开发环境、工程规范、提交与发布流程、如何新增 agent |
 | [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 |
 | [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md) | 贡献者行为准则 |

package/dist/agents/adapter.d.ts

@@ -80,7 +80,7 @@ export interface AgentRunResult {
     endReason?: string;
     /** GUI 实例是否因任务未真正完成而被保留 */
     keptInstance?: boolean;
-    needsUserKind?: "agent_question" | "close_existing_instance" | "login_required" | "system_permission" | "user_confirmation";
+    needsUserKind?: "agent_question" | "close_existing_instance" | "login_required" | "system_permission" | "setup_recovery" | "user_confirmation";
     pendingQuestion?: string;
     /**
      * 取消路径的 GUI 侧停止结果（GUI agent 专用）：

package/dist/agents/builtin.js

@@ -1,3 +1,4 @@
+import { ZCODE_SETUP_DEFAULTS } from "../config/schema.js";
 export const BUILTIN_PROFILES = {
     codex: {
         displayName: "Codex (ChatGPT 桌面端 GUI)",
@@ -27,6 +28,7 @@ export const BUILTIN_PROFILES = {
             scanPattern: "OpenAI.Codex_*_x64__*/app/ChatGPT.exe",
         },
         gui: {
+            ...ZCODE_SETUP_DEFAULTS,
             cdpPort: 9333,
             cdpPortAuto: true,
             cdpPortRange: 30,
@@ -73,8 +75,8 @@ export const BUILTIN_PROFILES = {
         authNote: "复用本机 ZCode 登录态",
         executableDiscovery: {
             dirs: [
-                "{ProgramFiles}/ZCode",
-                "{ProgramFiles(x86)}/ZCode",
+                "{PROGRAMFILES}/ZCode",
+                "{PROGRAMFILES(X86)}/ZCode",
                 "{LOCALAPPDATA}/Programs/ZCode",
                 "{LOCALAPPDATA}/ZCode",
                 "/Applications/ZCode.app/Contents/MacOS",
@@ -87,6 +89,7 @@ export const BUILTIN_PROFILES = {
             scanRoots: [],
         },
         gui: {
+            ...ZCODE_SETUP_DEFAULTS,
             cdpPort: 9333,
             cdpPortAuto: true,
             cdpPortRange: 30,
@@ -129,8 +132,8 @@ export const BUILTIN_PROFILES = {
         executableDiscovery: {
             dirs: [
                 "D:/TRAE Work CN",
-                "{ProgramFiles}/TRAE WORK CN",
-                "{ProgramFiles(x86)}/TRAE WORK CN",
+                "{PROGRAMFILES}/TRAE WORK CN",
+                "{PROGRAMFILES(X86)}/TRAE WORK CN",
                 "{LOCALAPPDATA}/Programs/TRAE WORK CN",
                 "{LOCALAPPDATA}/TRAE WORK CN",
                 "{APPDATA}/TRAE SOLO CN",
@@ -145,6 +148,7 @@ export const BUILTIN_PROFILES = {
             scanRoots: [],
         },
         gui: {
+            ...ZCODE_SETUP_DEFAULTS,
             cdpPort: 9222,
             cdpPortAuto: true,
             cdpPortRange: 20,

package/dist/agents/codex/run.js

@@ -1,3 +1,4 @@
+import { ZCODE_SETUP_DEFAULTS } from "../../config/schema.js";
 /**
  * Codex GUI 任务编排（开发计划 §4，对应需求 9 步）。
  *
@@ -41,6 +42,7 @@ const DEFAULT_DEPS = {
 export function codexGuiOf(resolved) {
     const g = resolved.profile.gui;
     return {
+        ...ZCODE_SETUP_DEFAULTS,
         cdpPort: g?.cdpPort ?? 9333,
         cdpPortAuto: g?.cdpPortAuto ?? true,
         cdpPortRange: g?.cdpPortRange ?? 30,

package/dist/agents/traework/run.js

@@ -1,3 +1,4 @@
+import { ZCODE_SETUP_DEFAULTS } from "../../config/schema.js";
 /**
  * TraeWork 单轮任务执行（开发计划 §4.5）。
  *
@@ -90,6 +91,7 @@ async function waitForUi(cdp, selectors, logger, timeoutMs = 60_000) {
 function guiOf(resolved) {
     const g = resolved.profile.gui;
     return {
+        ...ZCODE_SETUP_DEFAULTS,
         cdpPort: g?.cdpPort ?? 9222,
         cdpPortAuto: g?.cdpPortAuto ?? true,
         cdpPortRange: g?.cdpPortRange ?? 20,

package/dist/agents/zcode/cdp.d.ts

@@ -1,7 +1,7 @@
 import { CdpDisconnectedError, CdpUnavailableError } from "../traework/cdp/client.js";
 import { type ZcodeSelectorKey } from "./selectors.js";
 import type { ZcodePoll } from "./liveness.js";
-import type { ZcodeProjectItem } from "./project.js";
+import { type ZcodeProjectItem } from "./project.js";
 export { CdpDisconnectedError, CdpUnavailableError };
 export declare function retryZcodeEvaluation<T>(operation: () => Promise<T>, sleep?: (ms: number) => Promise<void>): Promise<T>;
 export interface ZcodeSessionItem {
@@ -14,6 +14,12 @@ export interface ZcodeQuestionAnswerResult {
     available: string[];
     error?: string;
 }
+export interface ZcodeClickExactResult {
+    clicked: boolean;
+    count: number;
+    available: string[];
+    testids?: string[];
+}
 export declare class ZcodeCdpClient {
     private readonly selectors;
     private readonly inner;
@@ -27,18 +33,19 @@ export declare class ZcodeCdpClient {
     exists(key: ZcodeSelectorKey): Promise<boolean>;
     text(key: ZcodeSelectorKey): Promise<string>;
     click(key: ZcodeSelectorKey): Promise<boolean>;
-    clickExact(key: ZcodeSelectorKey, value: string): Promise<{
-        clicked: boolean;
-        count: number;
-        available: string[];
-    }>;
-    selection(key: ZcodeSelectorKey): Promise<{
+    clickExact(key: ZcodeSelectorKey, value: string): Promise<ZcodeClickExactResult>;
+    selection(_key: ZcodeSelectorKey): Promise<{
         display: string;
         internal: string;
     }>;
     projects(): Promise<ZcodeProjectItem[]>;
     clickProject(id: string | undefined, projectPath: string | undefined): Promise<boolean>;
     boundProjectPath(): Promise<string>;
+    workspaceBinding(): Promise<{
+        triggerText: string;
+        projectPath: string;
+        ambiguous?: boolean;
+    }>;
     session(): Promise<{
         id?: string;
         title?: string;

package/dist/agents/zcode/cdp.js

@@ -1,6 +1,8 @@
 import { TraeworkCdpClient, CdpDisconnectedError, CdpUnavailableError, } from "../traework/cdp/client.js";
 import { candidateExpr } from "./selectors.js";
+import { projectDisplayName } from "./project.js";
 import { normalizeZcodeModelSelection } from "./model.js";
+import { projectTriggerDom, workspaceBindingExpression, modelSelectionExpression, sendButtonPointExpression, } from "./dom.js";
 export { CdpDisconnectedError, CdpUnavailableError };
 export async function retryZcodeEvaluation(operation, sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms))) {
     for (let attempt = 0;; attempt++) {
@@ -83,6 +85,13 @@ export class ZcodeCdpClient {
         return ((await this.evaluate(`(function(){for(const s of ${candidateExpr(key, this.selectors)}){for(const e of document.querySelectorAll(s)){const r=e.getBoundingClientRect();if(r.width&&r.height)return (e.value||e.textContent||e.getAttribute('title')||'').trim()}}return ''})()`)) || "");
     }
     async click(key) {
+        if (key === "projectTrigger") {
+            const point = await this.evaluate(`(function(){${projectTriggerDom(this.selectors)}if(!trigger)return null;const r=trigger.getBoundingClientRect();return {x:r.left+r.width/2,y:r.top+r.height/2}})()`);
+            if (!point)
+                return false;
+            await this.clickAt(point.x, point.y);
+            return true;
+        }
         const point = await this.evaluate(`(function(){for(const s of ${candidateExpr(key, this.selectors)}){for(const e of document.querySelectorAll(s)){const r=e.getBoundingClientRect();if(r.width&&r.height)return {x:r.left+r.width/2,y:r.top+r.height/2}}}return null})()`);
         if (!point)
             return false;
@@ -90,28 +99,42 @@ export class ZcodeCdpClient {
         return true;
     }
     async clickExact(key, value) {
-        const found = await this.evaluate(`(async function(){const norm=s=>(s||'').normalize('NFKC').trim().toLocaleLowerCase();const target=norm(${JSON.stringify(value)});const sels=${candidateExpr(key, this.selectors)};const visible=e=>{const r=e.getBoundingClientRect();if(!(r.width>0&&r.height>0&&r.bottom>0&&r.right>0&&r.top<innerHeight&&r.left<innerWidth))return false;const x=Math.max(0,Math.min(innerWidth-1,r.left+r.width/2)),y=Math.max(0,Math.min(innerHeight-1,r.top+r.height/2)),hit=document.elementFromPoint(x,y);return !!hit&&(hit===e||e.contains(hit))};const items=()=>{const out=[];for(const s of sels)for(const e of document.querySelectorAll(s))if(visible(e)&&!out.includes(e))out.push(e);return out};const leafTexts=e=>[...e.querySelectorAll('*')].filter(n=>n.children.length===0).map(n=>(n.textContent||'').trim()).filter(Boolean);const label=e=>(e.getAttribute('data-value')||e.getAttribute('data-model')||e.getAttribute('data-provider')||leafTexts(e)[0]||e.textContent||'').trim();let nodes=items();let scroller=nodes[0];while(scroller&&scroller!==document.body&&scroller.scrollHeight<=scroller.clientHeight)scroller=scroller.parentElement;const start=scroller?.scrollTop||0;if(scroller)scroller.scrollTop=0;const seen=new Map(),matches=new Map();for(let i=0;i<60;i++){await new Promise(r=>setTimeout(r,25));nodes=items();for(const e of nodes){const text=label(e);if(!text)continue;const id=e.getAttribute('data-id')||e.getAttribute('data-model-id')||e.getAttribute('data-value')||e.getAttribute('data-model')||e.getAttribute('data-provider')||e.getAttribute('data-testid')||text;seen.set(norm(id)+'|'+norm(text),text);if(norm(text)===target)matches.set(norm(id)+'|'+norm(text),{top:scroller?.scrollTop||0,text})}if(!scroller||scroller.scrollTop+scroller.clientHeight>=scroller.scrollHeight-1)break;const before=scroller.scrollTop;scroller.scrollTop=Math.min(scroller.scrollTop+Math.max(100,scroller.clientHeight*.8),scroller.scrollHeight);if(scroller.scrollTop===before)break}const matchesFound=[...matches.values()];if(matchesFound.length===1){if(scroller)scroller.scrollTop=matchesFound[0].top;await new Promise(r=>setTimeout(r,50));const exact=items().filter(e=>norm(label(e))===target);if(exact.length===1){exact[0].scrollIntoView({block:'center'});const r=exact[0].getBoundingClientRect();return {count:1,available:[...seen.values()],point:{x:r.left+r.width/2,y:r.top+r.height/2}}}}if(scroller)scroller.scrollTop=start;return {count:matchesFound.length,available:[...seen.values()]}})()`);
+        const found = await this.evaluate(`(async function(){const norm=s=>(s||'').normalize('NFKC').trim().toLocaleLowerCase();const target=norm(${JSON.stringify(value)});const sels=${candidateExpr(key, this.selectors)};const visible=e=>{const r=e.getBoundingClientRect();if(!(r.width>0&&r.height>0&&r.bottom>0&&r.right>0&&r.top<innerHeight&&r.left<innerWidth))return false;const x=Math.max(0,Math.min(innerWidth-1,r.left+r.width/2)),y=Math.max(0,Math.min(innerHeight-1,r.top+r.height/2)),hit=document.elementFromPoint(x,y);return !!hit&&(hit===e||e.contains(hit))};const items=()=>{const out=[];for(const s of sels)for(const e of document.querySelectorAll(s))if(visible(e)&&!out.includes(e))out.push(e);return out};const leafTexts=e=>[...e.querySelectorAll('*')].filter(n=>n.children.length===0).map(n=>(n.textContent||'').trim()).filter(Boolean);const label=e=>(e.getAttribute('data-value')||e.getAttribute('data-model')||e.getAttribute('data-provider')||leafTexts(e)[0]||e.textContent||'').trim();let nodes=items();let scroller=nodes[0];while(scroller&&scroller!==document.body&&scroller.scrollHeight<=scroller.clientHeight)scroller=scroller.parentElement;const start=scroller?.scrollTop||0;if(scroller)scroller.scrollTop=0;const seen=new Map(),seenTestids=new Map(),matches=new Map();for(let i=0;i<60;i++){await new Promise(r=>setTimeout(r,25));nodes=items();for(const e of nodes){const text=label(e);const testid=e.getAttribute('data-testid')||'';const id=e.getAttribute('data-id')||e.getAttribute('data-model-id')||e.getAttribute('data-value')||e.getAttribute('data-model')||e.getAttribute('data-provider')||testid||text;const key=norm(id)+'|'+norm(text);if(testid)seenTestids.set(key,testid);if(!text)continue;seen.set(key,text);if(norm(text)===target)matches.set(key,{top:scroller?.scrollTop||0,text})}if(!scroller||scroller.scrollTop+scroller.clientHeight>=scroller.scrollHeight-1)break;const before=scroller.scrollTop;scroller.scrollTop=Math.min(scroller.scrollTop+Math.max(100,scroller.clientHeight*.8),scroller.scrollHeight);if(scroller.scrollTop===before)break}const result={available:[...seen.values()],testids:[...seenTestids.values()]};const matchesFound=[...matches.values()];if(matchesFound.length===1){if(scroller)scroller.scrollTop=matchesFound[0].top;await new Promise(r=>setTimeout(r,50));const exact=items().filter(e=>norm(label(e))===target);if(exact.length===1){exact[0].scrollIntoView({block:'center'});const r=exact[0].getBoundingClientRect();return {count:1,...result,point:{x:r.left+r.width/2,y:r.top+r.height/2}}}}if(scroller)scroller.scrollTop=start;return {count:matchesFound.length,...result}})()`);
         if (!found.point)
-            return { clicked: false, count: found.count, available: found.available };
+            return {
+                clicked: false,
+                count: found.count,
+                available: found.available,
+                testids: found.testids,
+            };
         await this.clickAt(found.point.x, found.point.y);
-        return { clicked: true, count: found.count, available: found.available };
+        return {
+            clicked: true,
+            count: found.count,
+            available: found.available,
+            testids: found.testids,
+        };
     }
-    async selection(key) {
-        const raw = await this.evaluate(`(function(){for(const s of ${candidateExpr(key, this.selectors)})for(const e of document.querySelectorAll(s)){const r=e.getBoundingClientRect();if(!r.width||!r.height)continue;const data=e.closest('[data-model-id],[data-model],[data-value]')||e.querySelector('[data-model-id],[data-model],[data-value]')||e;return {display:(e.value||e.textContent||'').trim(),ariaLabel:(e.getAttribute('aria-label')||'').trim(),currentValue:(e.getAttribute('data-model-current-value')||'').trim(),legacyInternal:(data.getAttribute('data-model-id')||data.getAttribute('data-model')||data.getAttribute('data-value')||'').trim()}}return {display:''}})()`);
+    async selection(_key) {
+        const raw = await this.evaluate(modelSelectionExpression(this.selectors));
         return normalizeZcodeModelSelection(raw);
     }
     async projects() {
         return this.evaluate(`(function(){const out=[];for(const s of ${candidateExpr("projectItem", this.selectors)})for(const e of document.querySelectorAll(s)){const testid=e.getAttribute('data-testid')||'';const testPath=testid.startsWith('workspace-item-')?testid.slice('workspace-item-'.length):undefined;const name=(e.getAttribute('data-project-name')||e.querySelector('[class*=name]')?.textContent||e.textContent||'').trim();const p=e.getAttribute('data-project-path')||e.querySelector('[data-project-path]')?.getAttribute('data-project-path')||testPath||e.getAttribute('title')||undefined;const id=e.getAttribute('data-project-id')||e.getAttribute('data-id')||testid||undefined;if(name)out.push({name,path:p,id})}return out})()`);
     }
     async clickProject(id, projectPath) {
-        const point = await this.evaluate(`(async function(){const matches=[];for(const s of ${candidateExpr("projectItem", this.selectors)})for(const e of document.querySelectorAll(s)){const testid=e.getAttribute('data-testid')||'';const id=e.getAttribute('data-project-id')||e.getAttribute('data-id')||testid;const p=e.getAttribute('data-project-path')||e.querySelector('[data-project-path]')?.getAttribute('data-project-path')||(testid.startsWith('workspace-item-')?testid.slice('workspace-item-'.length):'')||e.getAttribute('title')||'';if((${JSON.stringify(id ?? "")}&&id===${JSON.stringify(id ?? "")})||(${JSON.stringify(projectPath ?? "")}&&p===${JSON.stringify(projectPath ?? "")}))matches.push(e)}if(matches.length!==1)return null;matches[0].scrollIntoView({block:'center'});await new Promise(r=>setTimeout(r,50));const r=matches[0].getBoundingClientRect();return r.width&&r.height&&r.bottom>0&&r.right>0&&r.top<innerHeight&&r.left<innerWidth?{x:r.left+r.width/2,y:r.top+r.height/2}:null})()`);
+        const displayName = projectPath ? projectDisplayName(projectPath) : "";
+        const point = await this.evaluate(`(async function(){${projectTriggerDom(this.selectors)}if(!trigger)return null;const targetName=norm(${JSON.stringify(displayName)});let containers=[];const controlledId=trigger?.getAttribute('aria-controls')||'';if(controlledId){const controlled=document.getElementById(controlledId);if(controlled&&visible(controlled))containers.push(controlled)}if(trigger){let parent=trigger.parentElement;while(parent&&!containers.length){if(parent.querySelector('[role="menuitemcheckbox"]'))containers.push(parent);parent=parent.parentElement}}if(!containers.length)containers=[...document.querySelectorAll('[role="menu"]')].filter(e=>visible(e)&&e.querySelector('[role="menuitemcheckbox"]'));const checkboxMatches=[];for(const container of containers)for(const e of container.querySelectorAll('[role="menuitemcheckbox"]'))if(visible(e)&&norm(e.getAttribute('aria-label')||e.getAttribute('data-value')||e.textContent||'')===targetName&&!checkboxMatches.includes(e))checkboxMatches.push(e);let selected=checkboxMatches.length===1?checkboxMatches[0]:null;if(!selected){const matches=[];for(const s of ${candidateExpr("projectItem", this.selectors)})for(const e of document.querySelectorAll(s)){const testid=e.getAttribute('data-testid')||'';const itemId=e.getAttribute('data-project-id')||e.getAttribute('data-id')||testid;const p=e.getAttribute('data-project-path')||e.querySelector('[data-project-path]')?.getAttribute('data-project-path')||(testid.startsWith('workspace-item-')?testid.slice('workspace-item-'.length):'')||e.getAttribute('title')||'';if(((${JSON.stringify(id ?? "")}&&itemId===${JSON.stringify(id ?? "")})||(${JSON.stringify(projectPath ?? "")}&&p===${JSON.stringify(projectPath ?? "")}))&&!matches.includes(e))matches.push(e)}if(matches.length===1)selected=matches[0]}if(!selected)return null;selected.scrollIntoView({block:'center'});await new Promise(r=>setTimeout(r,50));const r=selected.getBoundingClientRect();return visible(selected)?{x:r.left+r.width/2,y:r.top+r.height/2}:null})()`);
         if (!point)
             return false;
         await this.clickAt(point.x, point.y);
         return true;
     }
     async boundProjectPath() {
-        return ((await this.evaluate(`(function(){for(const s of ${candidateExpr("projectPath", this.selectors)})for(const e of document.querySelectorAll(s)){const p=e.getAttribute('data-project-path')||e.getAttribute('title')||(e.textContent||'').trim();if(p)return p}const trigger=document.querySelector('[data-testid="composer-workspace-trigger"]');const name=(trigger?.textContent||'').trim();if(name){const matches=[...document.querySelectorAll('[data-testid^="workspace-item-"]')].filter(e=>(e.textContent||'').trim()===name);if(matches.length===1)return (matches[0].getAttribute('data-testid')||'').slice('workspace-item-'.length)}return window.__ZCODE_PROJECT_PATH__||''})()`)) || "");
+        return (await this.workspaceBinding()).projectPath;
+    }
+    async workspaceBinding() {
+        return this.evaluate(workspaceBindingExpression(this.selectors));
     }
     async session() {
         return this.evaluate(`(function(){const stable=id=>id&&id!=='draft'?id:undefined;const idOf=e=>{if(!e)return undefined;const explicit=stable(e.getAttribute('data-session-id'));if(explicit)return explicit;const testid=e.getAttribute('data-testid')||'';return testid.startsWith('task-item-')?stable(testid.slice('task-item-'.length)):undefined};const selectors=['[data-testid^="v4-session-pane"][data-session-id]','[data-session-id][aria-current=true]','[data-session-id][aria-selected=true]','[data-testid^="task-item-"][aria-current=true]','[data-testid^="task-item-"][aria-selected=true]','[data-testid^="task-item-"][data-state=active]','[data-testid^="task-item-"][data-state=selected]'];let active=null;for(const s of selectors){active=[...document.querySelectorAll(s)].find(e=>{const r=e.getBoundingClientRect();return r.width&&r.height})||null;if(active)break}const id=idOf(active)||stable(history.state?.sessionId)||stable(history.state?.taskId)||undefined;const item=id?[...document.querySelectorAll('[data-testid^="task-item-"]')].find(e=>(e.getAttribute('data-testid')||'').slice('task-item-'.length)===id):null;return {id,title:id?(item?.getAttribute('title')||item?.textContent||active?.getAttribute('title')||'').trim().slice(0,240)||undefined:undefined}})()`);
@@ -141,20 +164,16 @@ export class ZcodeCdpClient {
         await this.send("Input.insertText", { text });
     }
     async sendMessage() {
-        if (!(await this.click("sendButton"))) {
-            await this.send("Input.dispatchKeyEvent", {
-                type: "rawKeyDown",
-                key: "Enter",
-                code: "Enter",
-                windowsVirtualKeyCode: 13,
-            });
-            await this.send("Input.dispatchKeyEvent", {
-                type: "keyUp",
-                key: "Enter",
-                code: "Enter",
-                windowsVirtualKeyCode: 13,
-            });
+        const deadline = Date.now() + 10_000;
+        for (let attempt = 0; attempt < 50 && Date.now() < deadline; attempt++) {
+            const point = await this.evaluate(sendButtonPointExpression(this.selectors));
+            if (point) {
+                await this.clickAt(point.x, point.y);
+                return;
+            }
+            await this.evaluate("new Promise(resolve=>setTimeout(resolve,200))");
         }
+        throw new Error("ZCode 发送按钮未在观察期内启用或被遮挡；未发送任务");
     }
     async answerQuestion(answer) {
         const located = await this.evaluate(`(function(){const norm=s=>(s||'').normalize('NFKC').trim().toLocaleLowerCase();const target=norm(${JSON.stringify(answer)});const visible=e=>{const r=e.getBoundingClientRect(),s=getComputedStyle(e);return r.width>0&&r.height>0&&r.bottom>0&&r.right>0&&r.top<innerHeight&&r.left<innerWidth&&s.display!=='none'&&s.visibility!=='hidden'};const boxes=[...document.querySelectorAll('[role="listbox"][aria-label]:has([role="option"])')].filter(visible);if(boxes.length!==1)return {count:boxes.length,available:[],buttonStates:[]};const box=boxes[0];const options=[...box.querySelectorAll('[role="option"]')].filter(visible);const label=e=>{const leaves=[...e.querySelectorAll('*')].filter(n=>n.children.length===0).map(n=>(n.textContent||'').trim()).filter(Boolean);return leaves.find(x=>norm(x)===target)||e.getAttribute('aria-label')||e.getAttribute('data-value')||''};const available=options.map(e=>{const leaves=[...e.querySelectorAll('*')].filter(n=>n.children.length===0).map(n=>(n.textContent||'').trim()).filter(Boolean);return leaves.find(x=>x&&!/^\\d+[.\uff0e]$/.test(x))||(e.textContent||'').trim()}).filter(Boolean);const matches=options.filter(e=>norm(label(e))===target);let scope=box,buttons=[];while(scope.parentElement){scope=scope.parentElement;buttons=[...scope.querySelectorAll('button:not([role="option"])')].filter(visible);if(buttons.length)break}if(matches.length!==1)return {count:matches.length,available,buttonStates:buttons.map(e=>e.disabled||e.getAttribute('aria-disabled')==='true')};matches[0].scrollIntoView({block:'center'});const r=matches[0].getBoundingClientRect();return {count:1,available,point:{x:r.left+r.width/2,y:r.top+r.height/2},buttonStates:buttons.map(e=>e.disabled||e.getAttribute('aria-disabled')==='true')}})()`);

package/dist/agents/zcode/dialog.d.ts

@@ -1,3 +1,9 @@
+export interface NativeDialogOptions {
+    timeoutMs?: number;
+    signal?: AbortSignal;
+    platform?: NodeJS.Platform;
+    onProgress?: (stage: string) => void;
+}
 export declare function parseMacSheetBaseline(baseline: string[]): number | null;
 export declare function validateMacSheetBaseline(baseline: string[]): {
     ok: true;
@@ -6,8 +12,8 @@ export declare function validateMacSheetBaseline(baseline: string[]): {
     ok: false;
     message: string;
 };
-export declare function listOwnedDialogs(pids: number[]): Promise<string[]>;
-export declare function selectZcodeFolder(folder: string, ownerPids: number[], baseline: string[]): Promise<{
+export declare function listOwnedDialogs(pids: number[], options?: NativeDialogOptions): Promise<string[]>;
+export declare function selectZcodeFolder(folder: string, ownerPids: number[], baseline: string[], options?: NativeDialogOptions): Promise<{
     ok: boolean;
     needsPermission?: boolean;
     message: string;