• CI failure tickets: only for main, and always deliverableIssue #1988 · PR #1990 · 2026-10-29 17:06 UTC+8
• health_alert_repo = "" turns off health-alert Issues, detection staysIssue #1985 · PR #1987 · 2026-10-29 09:53 UTC+8
• CI failure: tests on branch main (push)Issue #1983 · PR #1984 · 2026-10-28 23:28 UTC+8
• Delivery changes only what the ticket asks for, not unrelated files to get checks greenIssue #1980 · PR #1982 · 2026-10-28 22:36 UTC+8
• Clarification gate: propose a split whenever single_outcome fails, even with acceptance also missingIssue #1978 · PR #1979 · 2026-10-28 20:53 UTC+8
• Rename /orbi add-criteria to /orbi amendIssue #1975 · PR #1977 · 2026-10-28 20:11 UTC+8
• Clarification proposal and the criteria heading follow the repository's message languageIssue #1971 · PR #1974 · 2026-10-28 16:35 UTC+8
• Clarification gate can propose acceptance criteria: drop the 'stated in the text' rule for criteriaIssue #1967 · PR #1970 · 2026-10-28 14:59 UTC+8
• A stop comment with a proposal gives one instruction, not twoIssue #1962 · PR #1966 · 2026-10-28 13:41 UTC+8
• Make the progress-body outcome test independent of the clockIssue #1961 · PR #1965 · 2026-10-28 13:19 UTC+8
• Clarification gate: return the verdict through a schema-checked Pi tool so proposals are not droppedIssue #1955 · PR #1959 · 2026-10-28 11:16 UTC+8
• Journal a clarification proposal the gate drops instead of dropping it silentlyIssue #1958 · PR #1960 · 2026-10-28 10:57 UTC+8
• Reject milestone commands on stopped tickets without starving valid split commandsIssue #1939 · PR #1943 · 2026-10-28 08:49 UTC+8
• Preserve the outer seam patch when an inner MonkeyPatch context exitsIssue #1953 · PR #1954 · 2026-10-28 08:24 UTC+8
• CI failure: tests on branch main (push)Issue #1948 · PR #1952 · 2026-10-28 07:35 UTC+8
• Reject split retries when existing children differ from the edited proposalIssue #1940 · PR #1946 · 2026-10-28 05:41 UTC+8
• Ignore forged split completion receipts from other commentersIssue #1938 · PR #1942 · 2026-10-28 04:55 UTC+8
• /orbi add-criteria: add the proposed acceptance criteria and startIssue #1929 · PR #1941 · 2026-10-28 03:37 UTC+8
• /orbi split: turn the proposed split into child ticketsIssue #1927 · PR #1936 · 2026-10-28 01:56 UTC+8
• Preserve dependency targets when parsing edited proposal numbersIssue #1935 · PR #1937 · 2026-10-28 00:29 UTC+8
• Clarification stop comment proposes a split plan or acceptance criteriaIssue #1926 · PR #1934 · 2026-10-27 22:26 UTC+8
• Run registered ticket commands from the CLIIssue #1930 · PR #1933 · 2026-10-27 21:47 UTC+8
• One /orbi prefix for in-ticket commands, backed by a subcommand registryIssue #1922 · PR #1925 · 2026-10-27 10:56 UTC+8
• Stop with an approve-this-permission comment when the installation cannot read PR checks, instead of retrying silentlyIssue #1918 · PR #1921 · 2026-10-26 10:26 UTC+8
• Claim the longest-waiting ai-ready Issue first, not the newestIssue #1914 · PR #1917 · 2026-10-25 10:48 UTC+8
• CI failure: tests on branch main (push)Issue #1912 · PR #1913 · 2026-10-24 20:13 UTC+8
• refactor(release): extract release execution and closeout after the declaration parser movesIssue #1859 · PR #1911 · 2026-10-24 19:11 UTC+8
• fix(clarify): ignore stop markers quoted in runner diagnosticsIssue #1908 · PR #1909 · 2026-10-24 18:50 UTC+8
• Extract release declaration parsing from the release state machineIssue #1643 · PR #1910 · 2026-10-24 17:49 UTC+8
• refactor(review): split the 686-line _review_and_merge_round into named phasesIssue #1858 · PR #1906 · 2026-10-24 17:15 UTC+8
• refactor(review): extract merge execution and closeout from review_merge.pyIssue #1857 · PR #1907 · 2026-10-24 16:40 UTC+8
• fix(clarify): an unchanged ticket re-labelled ready is not stopped and commented on againIssue #1904 · PR #1905 · 2026-10-24 16:21 UTC+8
• Extract execution-source freshness preflight from the frozen runner moduleIssue #1650 · PR #1902 · 2026-10-24 14:35 UTC+8
• Extract review verdict parsing from review and merge orchestrationIssue #1644 · PR #1903 · 2026-10-24 14:00 UTC+8
• refactor(core): publicize cross-module helpers in checks, github, gitops and comment_trustIssue #1848 · PR #1901 · 2026-10-24 13:02 UTC+8
• refactor(core): one session-file finderIssue #1854 · PR #1899 · 2026-10-24 11:33 UTC+8
• refactor(core): one label-name parserIssue #1853 · PR #1900 · 2026-10-24 11:13 UTC+8
• fix(resume): a merged delivery PR whose Issue GitHub left open no longer freezes the queueIssue #1895 · PR #1897 · 2026-10-24 10:21 UTC+8
• refactor(core): one CI-pass classifier for the merge and release gatesIssue #1889 · PR #1898 · 2026-10-24 10:10 UTC+8
• Preserve review-key diagnostics after the provider config refactorIssue #1893 · PR #1896 · 2026-10-24 09:41 UTC+8
• refactor(core): one exact-milestone lookupIssue #1890 · PR #1894 · 2026-10-24 08:07 UTC+8
• refactor(core): publicize cross-module helpers in progress, failure, delivery_scene and pi_sessionIssue #1849 · PR #1891 · 2026-10-24 07:51 UTC+8
• refactor(config): return provider findings instead of a module attributeIssue #1852 · PR #1888 · 2026-10-24 06:19 UTC+8
• refactor(core): share one versioned comment-block parser across failure, scene and base-ci-redIssue #1855 · PR #1887 · 2026-10-24 05:18 UTC+8
• refactor(setup): extract config scaffolding and machine checks from pilot_setup.pyIssue #1861 · PR #1885 · 2026-10-24 04:57 UTC+8
• refactor(pi): extract idle recovery and stream logging from pi_process.pyIssue #1860 · PR #1886 · 2026-10-24 04:44 UTC+8
• Preserve the previous run ID when the test fixture's setter is patchedIssue #1882 · PR #1884 · 2026-10-24 04:13 UTC+8
• refactor(github): extract PR feedback and rollup helpers from github.pyIssue #1863 · PR #1877 · 2026-10-24 02:21 UTC+8
• refactor(cli): extract the argparse builder and subcommand handlersIssue #1862 · PR #1883 · 2026-10-24 02:10 UTC+8
• test(core): re-aim the patch ratchet and migrate one test file to FakeGh stateIssue #1865 · PR #1875 · 2026-10-24 00:48 UTC+8
• refactor(runner): extract PR verification into pr_verification.pyIssue #1864 · PR #1876 · 2026-10-23 23:42 UTC+8
• bug(tools): keep stdlib boundary stubs out of the internal patch ceilingIssue #1868 · PR #1873 · 2026-10-23 22:34 UTC+8
• bug(core): preserve config symlink targets during atomic milestone writesIssue #1872 · PR #1874 · 2026-10-23 21:39 UTC+8
• bug(tools): print current per-target patch counts in CIIssue #1869 · PR #1871 · 2026-10-23 20:57 UTC+8
• fix(core): make run-state, config and 429 state writes atomicIssue #1851 · PR #1870 · 2026-10-23 20:30 UTC+8
• Restrict base-CI recovery to the latest trusted failureIssue #1833 · PR #1835 · 2026-10-22 20:45 UTC+8
• Recognize concrete runtime failures in ticket preflightIssue #1831 · PR #1832 · 2026-10-22 20:21 UTC+8
• Reject a status_url whose authority has no hostnameIssue #1830 · PR #1834 · 2026-10-22 20:00 UTC+8
• base 分支 CI 已红（base_ci_red）：清晰评论、/orbi fix-base 与自动恢复Issue #1825 · PR #1829 · 2026-10-22 19:14 UTC+8
• Add a fixed project status link and readable summaries to task notificationsIssue #1826 · PR #1828 · 2026-10-22 18:01 UTC+8
• Keep Issue formatting instructions out of repository summary promptsIssue #1822 · PR #1824 · 2026-10-21 21:06 UTC+8
• Require evidence for exact file line citations in requirement draftsIssue #1819 · PR #1823 · 2026-10-21 20:48 UTC+8
• Remove invented line-count requirements from issue drafts and fidelity outputIssue #1816 · PR #1818 · 2026-10-21 14:11 UTC+8
• Delivery agent must not edit its own PR body: losing the run marker dead-loops resume into ai-blockedIssue #1813 · PR #1815 · 2026-10-20 13:58 UTC+8
• Redirect retired v0.1.1 docs links to the existing GitHub ReleaseIssue #1810 · PR #1812 · 2026-10-20 12:01 UTC+8
• Maintainers with private org membership are recognised from their repository permission, not the comment's authorAssociationIssue #1806 · PR #1809 · 2026-10-20 11:36 UTC+8
• Idle recovery terminates a hung command on hosts whose /proc has no stat file (subset=pid)Issue #1808 · PR #1811 · 2026-10-20 11:12 UTC+8
• A merge that landed while GitHub never marked the PR merged also closes the IssueIssue #1801 · PR #1805 · 2026-10-20 01:26 UTC+8
• The merge gate's landed verdict is used once; a later failed read cannot fail a landed mergeIssue #1802 · PR #1804 · 2026-10-20 00:47 UTC+8
• A merge that lands despite an error from the merge call ends as merged, not ai-fix-neededIssue #1798 · PR #1800 · 2026-10-19 22:39 UTC+8
• delivery_labels.py computes "has a delivery state" once: is_pickup_eligible calls has_delivery_stateIssue #1795 · PR #1796 · 2026-10-19 20:32 UTC+8
• claim.py decides raise-or-skip for a failed read in one functionIssue #1794 · PR #1797 · 2026-10-19 20:13 UTC+8
• The opened-PR scan finds the oldest free candidate even when older ones are skippedIssue #1792 · PR #1793 · 2026-10-19 19:43 UTC+8
• orbi status shows exactly the claim's decision: same held deliveries, and a failed release milestone read is a failed readIssue #1782 · PR #1791 · 2026-10-19 19:18 UTC+8
• Branch reclaim and the orphan-PR sweep use the one delivery-branch check: a non-Orbi branch is never deletedIssue #1788 · PR #1789 · 2026-10-19 18:41 UTC+8
• A CI failure never relabels a terminal Issue: the fix-needed label transition is the one decisionIssue #1786 · PR #1790 · 2026-10-19 18:28 UTC+8
• Label transitions clear the same queue label the claim uses, so a custom dispatch_label no longer survives a blocked terminalIssue #1775 · PR #1784 · 2026-10-19 17:18 UTC+8
• One check decides whether a PR is Orbi's delivery: triage and the merge notification agreeIssue #1776 · PR #1787 · 2026-10-19 16:09 UTC+8
• A failing review round reports the round it ran, not one re-derived from the scene it just advancedIssue #1783 · PR #1785 · 2026-10-19 15:42 UTC+8
• CI-failure triage reads the delivery state from the engine: a blocked delivery no longer gets ai-fix-neededIssue #1771 · PR #1778 · 2026-10-19 15:03 UTC+8
• The effective review round after a recovery is decided once: review and failure reports agreeIssue #1779 · PR #1781 · 2026-10-19 13:36 UTC+8
• orbi status shows as ready only an Issue the claim would pickIssue #1773 · PR #1780 · 2026-10-19 13:11 UTC+8
• Failure comments show the current PR's review round, the same one review usesIssue #1772 · PR #1777 · 2026-10-19 10:57 UTC+8
• Missing provider key, rate-limited release gate reads and gh network timeouts keep their pre-#1762 classificationIssue #1768 · PR #1770 · 2026-10-18 21:31 UTC+8
• Git errors are recognised the same way in any system languageIssue #1758 · PR #1767 · 2026-10-18 20:11 UTC+8
• Failed GitHub requests are classified by HTTP status, not by error textIssue #1760 · PR #1762 · 2026-10-18 19:52 UTC+8
• Chinese suggestion bodies use Chinese section headingsIssue #1764 · PR #1765 · 2026-10-18 19:14 UTC+8
• A failure that merely mentions "N failed" is no longer reported as a test failureIssue #1759 · PR #1763 · 2026-10-18 18:13 UTC+8
• A missing base branch is detected by asking the remote, not by matching Git's error textIssue #1756 · PR #1761 · 2026-10-18 17:30 UTC+8
• The missing-base-branch blocked comment says what is wrong, where to fix it and how to restartIssue #1753 · PR #1757 · 2026-10-18 15:48 UTC+8
• A failure in the first delivery step reaches the Issue instead of a silent retry every tickIssue #1749 · PR #1751 · 2026-10-18 13:51 UTC+8
• A delivery waiting for a maintainer no longer blocks reviews, resumes and new claimsIssue #1745 · PR #1752 · 2026-10-18 13:29 UTC+8
• Chinese repositories: the three spots #1739 left in English are Chinese tooIssue #1746 · PR #1750 · 2026-10-18 12:36 UTC+8
• Move bootstrap command and GitHub adapter tests into their own domain suiteIssue #1649 · PR #1743 · 2026-10-18 10:04 UTC+8
• Move bootstrap delivery and claim tests into their own domain suiteIssue #1648 · PR #1742 · 2026-10-18 09:31 UTC+8
• Move bootstrap configuration and preflight tests into their own domain suiteIssue #1647 · PR #1741 · 2026-10-18 08:26 UTC+8
• Chinese repositories still see English in Orbi's delivery comments and PR bodyIssue #1729 · PR #1739 · 2026-10-18 08:00 UTC+8
• Move bootstrap Pi session and recovery tests into their own domain suiteIssue #1646 · PR #1740 · 2026-10-18 06:54 UTC+8
• After the PR opens, the progress comment says it is waiting for checks and reviewIssue #1733 · PR #1735 · 2026-10-18 06:13 UTC+8
• Review scan takes the oldest open deliveries first, so the earliest PR is not starved by newer onesIssue #1734 · PR #1737 · 2026-10-18 05:46 UTC+8