calm-artist-7206
10/15/2026, 1:21 AMfresh-machine-37269
10/15/2026, 1:21 AMfresh-machine-37269
10/15/2026, 1:34 AMmap check --no-routes is ok on the committed map. The core change is sound β a known bug now only reads as xfail when the marked command fails and its # bug: PREDICATE matches, so an unrelated 401/404/503 can no longer stand in for a reproduction. I also confirmed the sibling ports (software-agent-sdk#5728, OpenHands#18348) are merged.
Three non-blocking nits in the review body:
1. assert --json FILE overloads a flag that means "JSON body string" on `api`/`exec`, where the file form is --json-file.
2. NEGATIVE_BUG_RE is searched over the whole command line, so a quoted JSON body or --check value containing ; if, && !, test ! or [ ! trips the "negative bug line" rejection.
3. # bug: exit eq 1 parses as a payload path (the payload really has an exit key) rather than the documented exit N form.
None of these block the merge β approval stands.calm-artist-7206
10/15/2026, 2:27 AMfresh-machine-37269
10/15/2026, 2:28 AMfresh-machine-37269
10/15/2026, 2:32 AMautomation-setup-draft-validity and automation-setup-draft-runs-error) swap raw text-red-300 for the theme token text-feedback-error, which resolves to #C0282D on Light+ and #BB2A26 on Solarized Light via --oh-feedback-error, so the pill goes from 1.49:1 to ~4.6:1 while dark themes keep their danger color. The earlier Validate PR description blocker is cleared on head ad6db62 (the linked issue now carries ready-for-dev), all other checks pass, and the SHA-pinned before/after Light+ captures cover the functional-evidence requirement. Mergeable state is now clean, with the approval in place.calm-artist-7206
10/15/2026, 2:33 AMfresh-machine-37269
10/15/2026, 2:37 AM0 9 * * *. The fix seeds customSchedule from the preset it replaces via the same toCron(...) used on save, with the guard value === "custom" && frequency !== "custom" so re-selecting Custom still preserves a typed cron, and loaded (editing) automations aren't clobbered.
I verified beyond the diff: on head a6f324d3, npm run make-i18n && npx vitest run __tests__/components/automations/automation-setup-panel.test.tsx passes 49/49, and removing only the new seeding block makes the added test fail (so the regression test is real, not vacuous). All acceptance criteria are met and the SHA-pinned before/after captures are present.
Both earlier blockers are now clear β Validate PR description (issue #18461 gained ready-for-dev) and `test-and-build (ubuntu)`/`(windows)` are all green β so both #18463 and #18464 are approved, and their mergeable state is now clean.
Note: per the repo's review rules I can only submit APPROVE or COMMENT, not REQUEST_CHANGES β if a human maintainer wants to block on anything, that call is theirs.