@mindstudio-ai/remy 0.1.87 → 0.1.88

package/dist/headless.js

@@ -413,6 +413,8 @@ ${isLspConfigured() ? `<typescript_lsp>
 
 <conversation_summaries>
 Your conversation history may include <prior_conversation_summary> blocks in the user's messages. These are automated summaries of earlier messages that have been compacted to save context space. The user does not see this summary, they see the full conversation history in their UI. Treat the summary as ground truth for what happened before, but do not reference it directly to the user ("as mentioned in the summary..."). Just continue naturally as if you remember the prior work.
+
+Old tool results are periodically cleared from the conversation to save context space. This is automatic and expected \u2014 you don't need to note down or preserve information from tool results. If you need to reference something from an earlier tool call, just re-read the file or re-run the query, or use your .remy-notes.md file.
 </conversation_summaries>
 
 <!-- cache_breakpoint -->

package/dist/index.js

@@ -5753,6 +5753,8 @@ ${isLspConfigured() ? `<typescript_lsp>
 
 <conversation_summaries>
 Your conversation history may include <prior_conversation_summary> blocks in the user's messages. These are automated summaries of earlier messages that have been compacted to save context space. The user does not see this summary, they see the full conversation history in their UI. Treat the summary as ground truth for what happened before, but do not reference it directly to the user ("as mentioned in the summary..."). Just continue naturally as if you remember the prior work.
+
+Old tool results are periodically cleared from the conversation to save context space. This is automatic and expected \u2014 you don't need to note down or preserve information from tool results. If you need to reference something from an earlier tool call, just re-read the file or re-run the query, or use your .remy-notes.md file.
 </conversation_summaries>
 
 <!-- cache_breakpoint -->

package/dist/subagents/codeSanityCheck/prompt.md

@@ -37,10 +37,14 @@ These are recurring mistakes the coding agent makes. If you see the conditions f
 
 - **Redundant userId on the auth table.** The auth table already has a platform-managed `id` column — that's the user ID. If the plan adds a `userId` column to the users/auth table, flag it.
 
-- **Too many granular API calls.** These apps are MVPs with small datasets. If the plan has separate method calls for every screen or sub-view (load profile, then load posts, then load post detail, then load comments), flag it. Favor fewer, fatter requests — a profile page that loads posts with full content means tapping a post is instant. A feed that includes comment previews and like state means the detail view renders from memory. Over-fetching at this scale is almost always the right call — users notice instant transitions, they don't notice a slightly larger payload.
+- **Too many granular API calls.** These apps are MVPs with small datasets. If the plan has separate method calls for every screen or sub-view (load profile, then load posts, then load post detail, then load comments), flag it. Favor fewer, fatter requests — a profile page that loads posts with full content means tapping a post is instant. A feed that includes comment previews and like state means the detail view renders from memory. Ideally there's one big batch call on first load that gets all the common data a user will need so subsequent navigation feels instant. Over-fetching at this scale is almost always the right call — users notice instant transitions, they don't notice a slightly larger payload.
 
 - **Wouter is not React Router.** The agent defaults to React Router patterns which silently break in wouter. Key differences: no `useNavigate()` (use `const [, setLocation] = useLocation()`), no `navigate(-1)` for back (use `window.history.back()`), no `element` prop on Route (use `component={Foo}` or children), no `<Routes>` (use `<Switch>` — without it all matching routes render simultaneously), no `<Navigate>` (use `<Redirect>`), no `<Outlet>` for nested routes (use `nest` prop on Route), and no `useSearchParams()` from react-router (wouter has its own version with a different setter API). If you see any of these React Router patterns in a wouter project, flag it.
 
+- **iOS mobile web touch/gesture handling.** If the plan involves horizontal swipe elements (carousels, image viewers, sliders) inside a scrolling page, flag these requirements: detect swipe direction by tracking both X and Y delta on touchmove, lock to horizontal after 8-10px of movement, only `preventDefault()` when locked horizontal, set `touch-action: pan-y` on the swipe container. On any draggable/swipeable container: `user-select: none`, `-webkit-user-select: none`, `-webkit-touch-callout: none`, `draggable="false"` on images, `pointer-events: none` on images inside the swipe track.
+
+- **Image preloading for detail views.** If the plan has a grid/list of thumbnails that link to detail views with full-size images, flag it if there's no preloading strategy. The fix: preload full-size images in the background (`new Image().src = url`) so they're in the browser cache by the time the user taps. This makes transitions feel instant.
+
 ## When to stay quiet
 
 Nits, style preferences, missing edge cases, things the agent will figure out as it goes, patterns that are "not ideal but fine," minor code smells. Let them slide. The agent is busy.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mindstudio-ai/remy",
-  "version": "0.1.87",
+  "version": "0.1.88",
   "description": "MindStudio coding agent",
   "repository": {
     "type": "git",