fix(agent): recover stale runs and improve new-user OA delivery

Fix DEV logout cookie clearing, materialize selected MindSpace OA assets before agent runs, and recover zombie runs from synced workspace pages. Add client run wait timeout, harness retry limits, page-edit asset forwarding, and logout/john2 scenario tests.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
john
2026-07-10 20:05:52 +08:00
parent aede1e6fcb
commit 7f8d692d16
34 changed files with 1318 additions and 125 deletions
+2 -1
View File
@@ -30,5 +30,6 @@ description: 网页抓取与搜索技能:访问网页、查阅文档、搜索
## 国内网络环境(建议)
- 生产环境访问不了 `google.com`,优先用本技能的 `web_search`DuckDuckGo/`fetch_url`,避免直接拿 `computercontroller__web_scrape` 抓 Google 页面来回重试
- `web_search` 连续几次没有可用结果时,改`fetch_url` 直接访问 `https://cn.bing.com/search?q=...``https://www.so.com/s?q=...` 这类国内可达的搜索入口
- **`web_search` 最多 2 次**(可换关键词);若仍无可用结果,**再 1 次**`fetch_url` 访问 `https://cn.bing.com/search?q=...``https://www.so.com/s?q=...`
- **3 轮搜索后仍无结果**:停止搜索,用内置知识直接生成页面/回答,并说明未获取实时搜索结果
- 百度/知乎/大众点评等站点有反爬拦截,遇到跳转或空结果就换个搜索源,不必在同一个来源上反复硬抓