Repository navigation
docs: T3 0.1.9-nightly.20261111.4515 review - #105
Merged
Merged
Conversation
Record evidence for pr-watch-is-native, delegate-task-reviews, scheduled-reviews, preview-tools-verification and rebuilt-worktree-no-setup. No status moves. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
T3 update review: 0.1.8-nightly.20261109.4429 → 0.1.9-nightly.20261111.4515 (10 releases)
209 release items, 11 watched files changed. No active entry breaks, no watched fix landed, and nothing is ready.
Needs you
agents/models.md: Claude Fable 5.2 (claude-fable-5-2) is now in the catalog (feat(models): add Claude Fable 5.2 pingdotgg/t3code#33500) andorchestrator_capabilitieslists it (efforts low…max, ultrathink; anultracodetoggle). No role names it. Decide whether to runevals/reviewer-recallon Fable 5.2 against Opus for the Claude side of Correctness and Standards/Spec. A model change goes through its own PR, which editsmodels.mdand the wrappers together.pr-watch-unreported-checks). Recommendation: leave it off.Improves
pr-watch-is-native: fix(server): stop PR watch flapping while workflows await approval pingdotgg/t3code#24599: checks waiting on a maintainer's workflow approval now wake once per head ("checks await approval") instead of flapping between failed and passed (awaitingApproval,approvalAnnouncedinthreadPullRequest.ts). #33548: the event stream described above. Evidence recorded. No skill edit:babysit-prreads the round from GitHub itself, and no project runs fork workflows.preview-tools-verification: fix(desktop): new browser tabs no longer lose waits, clicks and snapshots for good pingdotgg/t3code#32861 (closes #32848): "a fresh Browser panel tab … can stop answering everypreview_wait_for, locatorpreview_actandpreview_snapshotfor good, while main-worldpreview_evaluatestill works" (11 of 30 tabs failed, 16 of 30 under load). That matches the trial's lost clicks and the move topreview_evaluatefrom rubens feat(handoff): mark unchecked claims so the next agent does not build on them #72 on, though I haven't checked it against those threads' logs. Also #23589 (a tab the desktop shows readsvisible: trueand becomes the default tab), #33745 (snapshots keepvisibleTextbeside large accessibility trees) and #31925 (preview_evaluateaccepts top-levelawait). Evidence recorded. Retry when is still unmet because #23936 and #24812 are open. Proposed (yours to pick): when the next project PR needs a UI check anyway, run one realpreview_actdrive on this build without waiting for them. #32861 removes the failure mode that most looked like the trial's.delegate-task-reviews: #32504 (a follow-up task settles on a run its child queued on its own thread), #32815 (removing a queued completion keeps its siblings' wakes), #33446 (an unscopedt3 thread sendto a finished child wakes the parent; no skill uses it), #33295 (MCP calls read only the caller's active runs; result fields unchanged). Evidence recorded, nothing to adopt.scheduled-reviews: #32494: aschedule_taskupdate resets the workspace strategy only when the binding flips. Evidence recorded. Both reviews are unbound, so nothing changes for them.Drift
scheduled-run-worktree: the entry saysschedule_taskwith ascheduledTaskId"keeps it unlessbindToCurrentThreadis passed". Since fix(server): scheduled task updates keep the saved workspace unless the binding changes pingdotgg/t3code#32494 the stored strategy is kept unless the task moves between bound and unbound. Proposed edit in en-escena'sdocs/agents/t3-dependencies.md: "keeps it unless the update moves the task between bound and unbound (fix(server): scheduled task updates keep the saved workspace unless the binding changes pingdotgg/t3code#32494, since 0.1.9-nightly.20261111.4515; before, anybindToCurrentThreadreset it)". Its Breaks if still holds.main. The only tool name no entry spelled out wast3_preview_close(skills/implement/PREVIEW.mdstep 7, and each project'sverification.md).preview-tools-verificationcovers it and now names it.Not covered
None. infra and underthing have no
docs/agents/, so they are out of scope.Still ready
None.
Projects whose behaviour changes: none. This PR records evidence only; the en-escena edit and the Fable 5.2 decision above are proposals.
Review
Tier: readback, reviewed 9580ac5
recreateMissingWorktreeis the watch check that entry asks for; the preview inference already says it is unchecked against the trial logs); would merge: yes🤖 Generated with Claude Code