Repository navigation
Conversation
|
👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review. When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically. |
UX Review (Opus 5.5, fork) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing to check. [UX-REVIEWED] 1bad467 Description read[DESCRIPTION-READ] 92700fc29eabccb9c59ad471f546be2f9f9b08f320ad876146885083cad0adb8 That sha256 names everything this verdict read: the PR description, plus the 3 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Design Review (Opus 5.5, fork) — 🟡 CONCERNSDesign-level review of Design-Verdict: CONCERNS The harness asserts copy and a CI test that aren't on Watch
[DESIGN-REVIEWED] 1bad467 Description read[DESCRIPTION-READ] e87eef003fe0bc74df49a4248d29c30e96526243cafbf6bdd590c03f30c55207 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review (fork) — ✅ no blocking findingsReviewed Review detailsExits 1 on every run: the harness waits for @-menu and picker text that this SPA never renders. FINDING — website/scripts/capture-file-search-timeout.mjs:426 — In the current tree, [OPUS-REVIEWED] 1bad467 |
GPT 6.1 Review (fork) — ✅ no blocking findingsReviewed Review detailsNo findings. |
1a04387 to
ca842c3
Compare
ca842c3 to
6604335
Compare
6604335 to
43660b8
Compare
|
This capture harness photographs file-search failure/recovery UI copy ("File search timed out", "No access to the project folder", the picker/rail/folder-panel failure states) that is not on |
|
🤖 Kiro Crew [operator: dwu96#8c08bee0]: This PR has been inactive for 7+ days. I reviewed the blockers but they require your input:
One genuinely mechanical item, for when you pick that up:
When you've addressed these, the pipeline will re-assess on its next cycle. |
|
PR triage: OVERLAPPING. Keep open, continue development. Do not merge alone. I checked this diff (merge base What has changed in the stack since the last comments:
On main today, only the "No matching files" arm and the folder-listing copy from #7068 ("Folder listing timed out", "No access to this folder", "Unable to list folder", "Folder not found") can render. The harness still exits 1. Recommendation: re-scope the arms to what #7068 shipped plus #10838 and #10842. Either drop the #10839 arms or say which PR now owns them. Then land this with or after #10838 and #10842. The earlier GPT finding on |
Drives the built dashboard against a stubbed gateway to photograph each file-search and listing failure state, so the failure copy can be reviewed as rendered rather than read off the diff.
43660b8 to
1bad467
Compare
First Principles Review (Opus 5.5, fork) — ⏭️ skippedRevision |
|
Thank you @rnoack1 for this capture harness. Closing this because its subject is gone. The harness belongs to #10834, which was closed on 2026-10-06 as superseded by #7068. The copy it photographs ("File search timed out", "No access to the project folder") never landed: it is not in Your branch is kept. If that copy lands later, please reopen this or open a follow-up. |
Problem / Motivation
The dashboard's file-read surfaces — the
@-mention picker, the folder panel, the project and workspace pickers, the file rail — have a growing set of failure and recovery states: timed out, failed, no access, root missing, stale rows, and the retry affordances for each. None of them could be reviewed as rendered. A reviewer had to read the copy off the diff and take on trust that it appears where and when it is claimed to.There was no harness that drives those states, so "does this notice actually show up, with that wording, next to that button" was not an answerable question without hand-wedging a gateway.
Why it matters
Failure copy is exactly the category that is easy to get wrong in ways a unit test does not catch: a notice can assert correctly in jsdom and still be unreadable, mispositioned, or attached to a control the user cannot find. Two of the review lanes on the original change asked for rendered evidence and could not be satisfied, because producing it needed tooling that did not exist.
It also closes a specific trap: the harness serves the prebuilt bundle, so a copy change that has not been rebuilt does not appear. Reviewing a stale frame is worse than reviewing none, and having the driver in-tree is what makes the rebuild step explicit rather than remembered.
What changed (motivation → approach → change)
website/scripts/capture-file-search-timeout.mjs, a Playwright driver that serves the built dashboard and stubs the gateway per arm, then photographs each failure and recovery state in turn.lib/serve-dist.mjs,lib/stub-dashboard-api.mjs) rather than standing up its own server and stubs, so it behaves like the other capture scripts in this directory.This is tooling only. It adds no product code, imports nothing from
src/, and changes no behaviour: nothing in the shipped application calls it.Tests
None, deliberately: this is a developer capture script with no product surface, in a directory of sibling scripts that are likewise untested. Its correctness is observable directly — it either produces the frames or it does not.
Verified by execution rather than by assertion:
node --checkon the file,eslintclean, and the script runs to completion producing frames locally.Manual verification
eslintclean on the added file,node --checkpasses, and the harness runs end to end against a locally builtdist.One ordering note for a reviewer: several arms exercise states that arrive with the other units split out of the same original change. Landing the harness first is harmless — it is the tool, not the subject — but its full output is only meaningful once those merge. It is independent of them at build time and carries no prerequisite.