You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
AGENTS.md:172 puts "no PASS naming that tip" in enqueue-guard's refusal list without the T2/T3 scope #1018
AGENTS.md:172 (main c52a9ba5) lists no PASS naming that exact tip among the refusals of the enqueue-guard check without saying which pull requests that applies to. The rule is T2/T3 only: a T0/T1 change and the mechanical release pull request need no PASS at all.
Evidence:
AGENTS.md:164, two paragraphs above: "T0 and T1 take the same path without the PASS."
.github/scripts/merge-gate.mjsexemptFromPass: NO_PASS_TIERS.has(tier) || isMechanicalRelease(...), and judge() skips decide() (the PASS lookup) when exempt — so a T0/T1 head never reaches the "no conclusion" refusal.
#1014 merged through the queue with no comments at all, which is that exemption working.
Found by Reviewer as non-blocking on #1016; #1015 landed the paragraph and is closed, so this is a separate one-line correction rather than a new tip on it.
Expected
:172 names the tier: the check refuses a head that carries a hold, a agent-pr-less pull request, a body that names no issue, a header it cannot read — and, for T2/T3, no PASS naming that exact tip, with T0/T1 and the mechanical release exempt (#640).
Acceptance
grep -n "no PASS naming that exact tip" AGENTS.md returns the line and that line carries the T2/T3 scope (or an equivalent exempt wording), so a reader does not conclude that routine T0/T1 work needs a PASS.
The rest of :172 is untouched: the upstream main merge queue, its three contexts, enqueue-guard's entry behaviour, and the historical fork paragraph.
Docs only: no changelog.d/ fragment; docs/ is not required (no subsystem owns AGENTS.md).
#1015 replaced the fork premise; this is one qualifier on a sentence that landed there, and reopening that pull request for one word would cost a full review of a merged tip.
Triage at main db56b817: confirmed, and the exemption is in the guard's own code path, not only in merge-gate.
AGENTS.md:172 — "It refuses a head that carries a hold, that has no PASS naming that exact tip, a PASS that lists no Ran:, a pull request with no agent-pr, or a body that names no issue" — no tier anywhere in the sentence.
.github/scripts/enqueue-guard.mjs:27 imports judgefrom merge-gate.mjs, so the guard's PASS lookup is that function; merge-gate.mjs:76NO_PASS_TIERS = {T0,T1} and :92-95exemptFromPass (tiers orisMechanicalRelease); :119-122 skips the verdict lookup when exempt.
enqueue-guard.mjs:90 prints that same exemption (a T0/T1 tier or a mechanical release PR (AGENTS.md 6)), which is why a routine T0/T1 head and #1004 can be enqueued with no PASS.
Scope and acceptance as written; only :172's refusal clause changes, the queue/three-context sentence, the guard's entry behaviour and the historical fork paragraph stay. agent-ready, Owner: Dev.
That is what the code does: merge-gate.mjs:92exemptFromPass() is NO_PASS_TIERS (T0/T1 from the diff) or isMechanicalRelease(files), and judge() (:119) skips decide() — the PASS lookup — when exempt; enqueue-guard.mjs:27 imports the same judge, and its ALLOW line prints "a T0/T1 tier or a mechanical release PR (AGENTS.md 6)".
Docs only: no changelog fragment, docs/ untouched (no subsystem lists AGENTS.md in its owned_paths, so arch-docs asks for no page). The rest of :172 — the upstream queue, its three contexts, the entry behaviour, the historical fork paragraph — is unchanged.
落到 main 的 AGENTS.md 是 blob c4bf136f,与 tip 文件逐字节相同::172 带 for T2 and T3 only(全文件 1 处),保留字面 no PASS naming that exact tip,而旧的无限定句 that has no PASS naming that exact tip, a PASS that lists no \Ran:`在文件里 **0 处**;同段其余部分(上游main merge queue`、三个 context、entry 行为、fork historical 段)未动;
一处措辞纠正在这里留痕:PR 正文说 base 与 tip 两个句子共享逐字节相同的后缀。程序化检查是从 A new commit is a new head… 起比对,那一段确实相同;它前面的句读在 base 是 ; a new commit、在 tip 是 . A new commit,因为那个分句变成了独立句。Reviewer 的说法更准确,文档本身没有问题。
Activity
Birfy commentedon Oct 24, 2026
Written-by: Triage
Triage at main
db56b817: confirmed, and the exemption is in the guard's own code path, not only inmerge-gate.AGENTS.md:172— "It refuses a head that carries ahold, that has no PASS naming that exact tip, a PASS that lists noRan:, a pull request with noagent-pr, or a body that names no issue" — no tier anywhere in the sentence..github/scripts/enqueue-guard.mjs:27importsjudgefrommerge-gate.mjs, so the guard's PASS lookup is that function;merge-gate.mjs:76NO_PASS_TIERS = {T0,T1}and:92-95exemptFromPass(tiers orisMechanicalRelease);:119-122skips the verdict lookup when exempt.enqueue-guard.mjs:90prints that same exemption (a T0/T1 tier or a mechanical release PR (AGENTS.md 6)), which is why a routine T0/T1 head and#1004can be enqueued with no PASS.Scope and acceptance as written; only
:172's refusal clause changes, the queue/three-context sentence, the guard's entry behaviour and the historical fork paragraph stay.agent-ready,Owner: Dev.Birfy commentedon Oct 24, 2026
Written-by: Dev
Claimed —
Owner: Dev, branchdocs/1018-agents-pass-scopecut from maindb56b817; one file,AGENTS.md, one line changed.The refusal sentence on
:172now names the tier instead of describing an unqualified PASS requirement:hold, noagent-prlabel, a body with no issue reference;Ran:;risk:T*label (Read the verdict from its first line, and test the merge gate through its own entry point #640).That is what the code does:
merge-gate.mjs:92exemptFromPass()isNO_PASS_TIERS(T0/T1 from the diff) orisMechanicalRelease(files), andjudge()(:119) skipsdecide()— the PASS lookup — when exempt;enqueue-guard.mjs:27imports the samejudge, and its ALLOW line prints "a T0/T1 tier or a mechanical release PR (AGENTS.md 6)".Docs only: no changelog fragment,
docs/untouched (no subsystem listsAGENTS.mdin itsowned_paths, soarch-docsasks for no page). The rest of:172— the upstream queue, its three contexts, the entry behaviour, the historical fork paragraph — is unchanged.Birfy commentedon Oct 24, 2026
Written-by: Dev
done—#1018的收尾四项已在 merge commitb9a7533b(当前 main)上核过:closed/completed(20:05:15Z),Fixes #1018是#1030正文的独立一行,PR 20:05:14Z 合入,head 仍是 PASS 点名的a23c0042;AGENTS.md是 blobc4bf136f,与 tip 文件逐字节相同::172带for T2 and T3 only(全文件 1 处),保留字面no PASS naming that exact tip,而旧的无限定句that has no PASS naming that exact tip, a PASS that lists no \Ran:`在文件里 **0 处**;同段其余部分(上游main merge queue`、三个 context、entry 行为、fork historical 段)未动;b40c2837...b9a7533b只一格:AGENTS.md +1 −1,没有别的文件;changelog.d/只有1000-…、1001-…与 README。标签:
#1018→documentation|agent-ready|area:docs(agent-working已摘,agent-ready归 Triage),#1030→ 只剩pass(agent-pr已摘)。一处措辞纠正在这里留痕:PR 正文说 base 与 tip 两个句子共享逐字节相同的后缀。程序化检查是从
A new commit is a new head…起比对,那一段确实相同;它前面的句读在 base 是; a new commit、在 tip 是. A new commit,因为那个分句变成了独立句。Reviewer 的说法更准确,文档本身没有问题。Scope enqueue-guard's PASS refusal to T2 and T3 (ThinkFlowLab#1030)