fix(agent): solve_challenge authorization covers its child's coordinate steps

With only a browser.solve_challenge grant, the vision child run was
denied browser.click_at, so the loop stopped after the first screenshot
(REAL-2 task 4). Steps inside a run spawned by solve_challenge are now
allowed; ancestors' allowlist and step budget still apply. Needs image
rebuild.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
This commit is contained in:
rootandClaude Opus 5.5 committed 2026-09-29 03:06:05 +08:00
1 parent bd0e011748
commit 7dc7b1c5fa
3 files changed
+38 -1

No files matched your search

+1 -1
View File
@@ -136,7 +136,7 @@ execute 类需要持续授权(`automation_grant`)或审批。元素引用只来
- **证据**:每次截图(含动作后的截图)写 480 像素宽的 JPEG 缩略图到 `DATA_DIR/tmp/evidence/<id>.jpg`(0600,7 天后清理);
能力输出只给引用 `evidence/<id>.jpg`,随工具结果摘要进 `agent_run_event`。图片数据不进 JSON 文本、不进操作账本。
- **授权**:`click_at / type_at / scroll / solve_challenge` 由 `browser.interact` 的持续授权覆盖(也可单独授权);
父会话白名单里有 `browser.solve_challenge` 即覆盖子流程用到的截图与坐标能力。坐标动作按调用区分幂等键,崩溃恢复不重放。
父会话白名单里有 `browser.solve_challenge` 即覆盖子流程用到的截图与坐标能力;授权同理,solve_challenge 通过授权后,子流程(`spawn_reason = browser.solve_challenge`)内的坐标动作不再单独要持续授权。坐标动作按调用区分幂等键,崩溃恢复不重放。
- 交给模型的截图参数可在 `browser_provider.config` 调:`screenshotMaxWidth`(≤1280)、`screenshotFormat`(`jpeg` / `png`)、
`screenshotQuality`(20–95)。