[qdaxb]

问题

Wework 云端桌面 E2E 偶发在首次打开项目终端时失败。远端 executor 返回 Project path is outside allowed workspace roots，导致终端卡片进入“启动失败”状态，测试最终以等待终端窗口超时结束。

根因

测试创建的云端项目 workspace 位于临时结果目录中，但远端 executor 只通过 LOCAL_WORKSPACE_ROOT 间接获得较宽泛的目录边界，没有使用 executor 为额外工作区提供的显式 allow-list 配置。环境状态异常时，session 路径校验可能无法确认该 workspace。

同时，通用快照超时诊断没有保留 workspace 面板相关的 test ID，失败日志只显示 composer 控件，掩盖了实际的终端启动失败状态。

修改

• 启动测试用远端 executor 时，通过 WEGENT_WORKSPACE_ROOTS 显式声明本次云端 workspace。
• 快照超时诊断保留 workspace-* 和 bottom-workspace-* test ID，便于直接识别终端、launcher 和底部面板状态。

影响

云端桌面 E2E 的终端路径边界配置变得确定，不再依赖隐式目录覆盖；若终端流程再次失败，CI 日志也会包含直接相关的 UI 状态。

验证

• node --check wework/e2e/desktop/task-flow.e2e.mjs
• pnpm --filter wework exec prettier --check e2e/desktop/task-flow.e2e.mjs
• pnpm --filter wework exec eslint e2e/desktop/task-flow.e2e.mjs
• pnpm --filter wework e2e:desktop:cloud
  • 18 个模型协议组合全部通过
  • 云端项目创建、首次终端启动、历史终端恢复及任务流程通过
• pre-push Wework ESLint、TypeScript 和全量单测通过

Summary by CodeRabbit

• Bug Fixes
  • Desktop end-to-end diagnostics now retain workspace-related test identifiers in snapshots.
  • Cloud execution now consistently uses the configured workspace path.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 42234a0c-5605-478a-bea3-0b99d724b092

📥 Commits

Reviewing files that changed from the base of the PR and between 1edb23c and b72b463.

📒 Files selected for processing (1)

• wework/e2e/desktop/task-flow.e2e.mjs

---

📝 Walkthrough

Walkthrough

The desktop E2E flow now retains workspace-related test IDs in snapshot diagnostics and sets WEGENT_WORKSPACE_ROOTS for the cloud executor.

Changes

Workspace E2E flow

Layer / File(s)	Summary
Workspace diagnostics and executor configuration wework/e2e/desktop/task-flow.e2e.mjs	Snapshot diagnostics include workspace- and bottom-workspace- test IDs. The cloud executor environment sets WEGENT_WORKSPACE_ROOTS to the configured workspace path.

Estimated code review effort: 1 (Trivial) | ~5 minutes

Possibly related PRs

• wecode-ai/Wegent#1884: Extends the desktop E2E test with workspace-related diagnostics and executor configuration.
• wecode-ai/Wegent#2056: Directly relates to the same task-flow.e2e.mjs workspace changes.
• wecode-ai/Wegent#2074: Updates the same desktop cloud E2E flow and workspace setup.

Suggested reviewers: earthalone2026, sdadunderscoresdad

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes hardening the Wework cloud terminal E2E workspace root handling.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

• Create stacked PR
• Commit on current branch

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch agent/harden-cloud-terminal-e2e

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}