Files
memind/docs/regression-guards/README.md
T
john ef4ce12bbf fix(vision): stop read_image from poisoning text-provider sessions
Image turns already get a vision-model description injected into the prompt,
but the agent kept calling read_image to "confirm" the pictures. Those tool
results carry base64 image parts that Goose persists, so every later turn
against the text-only chat provider failed with `unknown variant image_url`
before the agent could write the page. WeChat page requests therefore fell
through to the fail-closed delivery message.

Drop read_image for the turn whenever a vision model handles the images, say
so explicitly in the injected prompt, and teach the poison scan to recognise
tool image parts so already-polluted sessions rotate instead of failing again.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-22 10:40:12 +08:00

2.9 KiB
Raw Blame History

回归守卫索引

本目录记录已验证、不可轻易回退的行为。适用于 Codex、Cloud、Cursor 及任何人工作前的自检。

生产发布的跨模块完整场景和硬闸门不在本目录重复维护,统一以 生产发布守门员 为准;本目录中的回归守卫必须映射并纳入该发布守门员。

文档

文档 保护内容
mindspace-publish-and-chat-finish.md edit_file 覆盖 public/*.html ② Finish 后聊天不清空、不暴露 agent 内部前缀
mindspace-remote-page-sync-and-thumbnail.md ① remote 模式 public HTML 入库 sync ② storage 缺失时缩略图/读页回退 workspace HTML
page-data-delivery-contract.md 数据集注册、绑定、真实 page UUID 与交付验收
h5-session-stream-replay.md Portal/Goose SSE 游标映射、断线续播与 Finish 终态恢复
memory-v2-candidate-and-lifecycle.md 候选记忆表幂等初始化、Portal fail-open、生命周期 off/canary/active 作用域
episodic-history-recall.md 历史会话召回、用户隔离、旧快照回退、提示注入与 off/canary/active 灰度
mindspace-seo-geo.md 公开页 SEO/GEO 注入、私有页 noindex、sitemap/llms.txt 与百度推送开关
vision-turn-read-image-isolation.md 图片轮次禁用 read_image、视觉提示硬约束、read_image 工具图片污染检测与会话轮换

自动化

npm run verify:mindspace-publish-guards        # 推荐:改相关代码后
npm run verify:mindspace-publish-guards:full   # 发 Portal runtime 后
npm run verify:h5-session-patches              # H5 会话、Agent Run 与 SSE 续播
npm run verify:seo-geo                           # SEO/GEO 单测
npm run verify:seo-discovery                     # robots/sitemap/llms 路由

新增守卫时

  1. 在本目录添加说明 md(症状 + 必须保留的行为 + checklist
  2. 添加/扩展 *.test.mjs 并接入 npm test
  3. scripts/verify-*.mjs 增加源码片段检查(如适用)
  4. 更新根目录 AGENTS.md 索引
  5. (可选)在 .cursor/rules/ 添加对应 .mdc 供 Cursor 自动提示

工具可见性

机制 Cursor Codex / Cloud / 其它
本目录 + AGENTS.md (需读仓库文件)
npm run verify:*
npm test
发版脚本门禁
.cursor/rules/*.mdc 自动 不自动

其它 AI 工具不会自动读 Cursor 规则;请以 AGENTS.md 与本目录为准。