Conversation
The Script body opened at /studio?demo=1 while judge step 2 says to open / and call get_site_capabilities, so the written script contradicted the judging instructions and never demonstrated the two tools shipped in v2.28.0. New 0:00-0:22 opening: the agent calls get_site_capabilities, reads the four-route tool map and the boundaries field before any tool runs, calls find_profile, and follows the demoStudio entry point into the Studio itself. The 1:30 segment now opens the URL find_profile returned rather than a typed address, and the architecture card gains the landing row. Rebudgeted rather than overran. The script was already at the 2:50 cap, so Studio co-design goes 70s to 68s: 22 + 68 + 50 + 30 = 170s = 2:50, contiguous, still under the three-minute limit. Renaming the Studio heading to 0:22 broke the anchor at webmcp.md:88; fixed. webmcp.md:80's landing role-play link pointed at the judge checklist rather than a role-play, because no landing segment existed to link to; it now points at the real segment. All four anchors verified against the actual headings. The 19 TODO_ placeholders are deliberate and untouched. Fixes #1254 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QeMnjM7oeVWTbGwHbs5cfr
Verified v2.28.0 in production with flagged Chrome. All 19 registrations across 18 distinct names are live and execute correctly, but the surfaces do not register at the same speed. Measured 2026-09-01 against production: /studio?demo=1 had all 9 tools at 0ms, while /u/:handle reported zero tools 8.7 seconds after navigation and then registered all 6 once hydration completed. The share page is heavier and its tool host mounts after hydration. An agent or judge that lists tools immediately on navigation therefore sees an empty list, which reads as broken. Recorded in the 1:30-1:48 segment and in judge instruction 7, the two places a reader acts on it. No timing change: the four segments still total 2:50. Refs #1254 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QeMnjM7oeVWTbGwHbs5cfr
The assertion at StudioClient.render.test.tsx sampled aria-busy once, outside any waitFor. The waitFor above it gates on the terminal line count, which "initial 2 + input + result" can satisfy before the save promise resolves, so the preview pane is still aria-busy="true" at that moment. On slower CI runners it failed with "expected 'true' to be 'false'". It failed a required check during the v2.28.0 release and cost a rerun cycle. Reproduced deterministically before fixing, rather than assumed: holding the race window open by resolving the fetch mock on a 15ms timer reproduced the exact CI assertion locally. Two changes. The assertion now waits on its own condition, matching the sibling "Saving..." test 50 lines above which already used that pattern. And the fetch mock keeps resolving on a real timer, so the race window stays open on every run. That converts a flaky test into a deterministic one: removing the waitFor now fails 3 runs out of 3 rather than intermittently on CI. Verified: 8 consecutive passing runs of the file, then 494 files / 8231 tests, typecheck and lint clean. Fixes #1253 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QeMnjM7oeVWTbGwHbs5cfr
All values probed against production or GitHub this session: live URL, handle juan294, verification hash 84567a48...a57, public repo URL, release SHA 4d8a6b7 (v2.28.0), flag evidence. Gates 1 to 3 closed. Release shipped and verified today. The flag flip needed no action, all three were already enabled on 2026-08-27, and the live catalog was confirmed in flagged Chrome at 19 registrations across 18 names. The repo is already public with the root LICENSE detected as MIT. Corrects the 2:05 beat: the altered-hash page returns 200 and renders an unverified state, only /api/verify answers 404. Expecting a browser error page on camera would be wrong. TODO_YOUTUBE_URL and TODO_DEVPOST_SUBMISSION_URL remain, both Juan-gated. The verify step now expects exactly those two rather than zero. Refs #1254 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QeMnjM7oeVWTbGwHbs5cfr
Contributor
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Bundle Size Report
|
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Verification
Operations
Migration 038 seeds the remote MCP flag as disabled. Applying that migration and enabling the production flag require separate authorization.