Repository navigation
[AGENT] Select browser CI separately from visual evidence - #370
Conversation
PR verification reportAll required checks and selected coverage passed. Browsers selected for this run: yes.
Changed visual baselines: none. Updated from Baseline Checks run 38686489081, attempt 1 for Selection observed at 2026-10-15T08:37:35.353Z for commit |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
[AGENT] Claude ReviewClaude review for commit This is not a current review or approval. |
|
[AGENT] Disposition for the Claude review’s nonblocking duplication note: retained. The dependency-prepared core job runs all three contract files, including the YAML-backed removal tests. The report job repeats the small dependency-free helper checks and still runs via |
 [High impact] The reviewed changes appear safe to merge; no actionable defects were found. SummarySeparates browser coverage from visual evidence while keeping core checks and the web build on by default.
Review checks were static; no tests or workflows were run during this review. Diagram%%{init: {'theme': 'neutral'}}%%
flowchart TD
Event[Normal PR event] --> Selection[Snapshot labels]
Event --> Core[Core checks and web build]
Selection --> Mode{Selected coverage}
Mode -->|Neither label| Skip[Skip browsers and Storybook]
Mode -->|ci:browsers| Browsers[Browser assertions and Storybook build]
Mode -->|ci:visuals| Browsers
Mode -->|ci:visuals| Visuals[Review captures and both comparisons]
Core --> Report[Check results and selected artifacts]
Skip --> Report
Browsers --> Report
Visuals --> Report
Report --> Head{PR head still matches?}
Head -->|Yes| Publish[Update report with run selection]
Head -->|No| Ignore[Do not publish stale evidence]
Reviews (2) · Last reviewed commit: "Preserve diagnostics and execute CI sele..." · Reviewed by Greptile |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
[AGENT]
Default PR CI runs core checks and the web build without browsers or a Storybook build.
ci:browsersselects all existing browser assertions and the Storybook build.ci:visualsalone implies browsers and also selects review captures and both image comparisons.Selection is fixed at the normal PR event. Label edits do not launch application CI. The report validates core results, selected coverage and current artifacts, marks unselected coverage explicitly, and fails on unknown selection, unexpected skips and stale evidence. Local defaults, functional diagnostics, assertions, baselines and deploy controls are preserved. Agent and developer docs describe when to select each mode.
All three Linux runs passed on
45ac074af8efb79f0b5874f79ac23a26afdbf409:These are sums of elapsed job durations, including parallel jobs, for Baseline Checks only. They are not wall time, billed dollars or guaranteed savings. The earlier reference run used 40.5 summed job minutes. No baselines changed.
The downloaded browsers-only local-flow artifact retained the profile, search and submit-success diagnostic PNGs. Linux full scripts passed with 261 passed, zero failed and one skipped. All 21 CJS selection/report/removal contracts passed on Linux, including exact inline trusted code and stale-head/remove-readd races. Local focused contracts passed 34/34, and full docs verification passed. The unchanged Windows Gemini shutdown timing failure remains outside this change: latest local full run was 260 passed, one failed and one skipped.
Independent whole-branch review found two important issues, both reproduced and fixed before publication. Current-head Claude and Greptile reviews found no actionable defects. Codex and Copilot reviews are quota-unavailable, not approvals. The nonblocking duplicate-helper observation is disposed here.
The existing live ci:visuals removal run passed, preserving prior results and appending the historical note without launching application CI. Since
pull_request_targetruns trusted base code, the bounded liveci:browsersremoval check remains pending after merge under the accepted evidence boundary. Its exact inline contracts passed on Linux; trust controls were preserved.No merge or deployment is included.