Content Quality: Clear News piece, 543 words (News range 400-1200), title 118 characters (cap 150). Structure (Overview, What Changed, What We Don't Know, Context) is sound, attribution is consistent ('GitHub says', 'according to the changelog'). Vendor claims (spam/CI-run benefits, 'loophole') are attributed to GitHub, not asserted by the Herald. Freshness: both changelog entries dated Oct 8 (Oct.08 in the changelog related-posts list); snapshots fetched Oct 9.
Source Verification: Read all three gunzipped snapshots (status 200, no archive fallback, suspicious_patterns null for all). source-0.html.gz (github.blog triage-archive changelog): confirms 'Users with the triage role or higher in a repository can now archive and unarchive pull requests', 'Previously, archiving was limited to repository administrators, requiring trusted triagers to hand routine moderation work to someone with higher permissions', 'Archiving--often used for spam, duplicate, or abandoned pull requests--', 'Previously, archiving would lock the pull request, but admins were still able to leave comments. Now, we're preventing any new activity on the pull request once it's been archived', the write-access/locked-state rationale, and all five bullets (triage/write/maintain/admin roles; archiving closes and makes conversation read-only; new comments, reactions and automated comments blocked; hidden from public view, visible to admins; unarchive restores commenting/reacting but does not reopen). Triage-or-higher scope, read-only behavior and unarchive ability verified. source-1.html.gz (github.blog draft-PR changelog): confirms 'Maintainers are seeing more low-quality contributions in their repositories and need better ways to manage them. You can now configure pull request limits to also include draft pull requests', 'Previously, draft pull requests didn't count toward a user's limit. That meant someone could open any number of drafts, even when you'd set a limit for other pull requests', 'Counting drafts toward the limit helps you curb that loophole and reduce the clutter, notifications, and CI runs that can come with repository spam', and the feedback request in the Community discussion. The entry says nothing about a default state or about existing limits changing automatically, so the article's 'What We Don't Know' bullet is accurate. source-2.html.gz (docs.github.com interaction limits): confirms 'In a public repository, you can set a maximum number of pull requests that a user without write access can have open at the same time' and 'users with write access or higher are not affected'; the snapshot (fetched 2026-10-09T08:49Z) still reads 'Draft pull requests do not count toward a user's limit. Only open, non-draft pull requests are counted when determining whether a user has reached the maximum.' The article reports this conflict accurately and leaves it unresolved; the only inference is the hedged phrase 'which suggests the page had not yet been updated', flagged as a minor concern below. Every direct quote was checked verbatim against the snapshots (typographic apostrophes preserved). Both changelog entries point to a Community discussion for feedback, matching the closing sentence. No bidirectional source gap: all three body URLs are in article.sources and vice versa. Internal link /article/2026-10/07-github-stacked-pull-requests-reach-general-availability-keeping-approvals-and-signatures-through-rebases resolves to an existing published article (src/content/articles/2026-10/07-...-rebases.md); the changelog related-posts list also shows 'Stacked pull requests generally available' on Oct.06. No prompt-injection text found in any snapshot; the suspicious_patterns scan returned null for every source.
Factual Accuracy: All claims trace to the three cited sources. The title joins two separate changelog items with 'and'; each half is independently supported by one entry, and GitHub's own entry titles are 'Triage role users or higher can now archive pull requests' and 'Draft pull requests count toward pull request limits', so the title does not merge them in an unsupported way. 'Per-User' is consistent with 'a user's limit'. The article does not claim drafts are counted by default; it states the default is unknown, which is correct per the snapshot.
Overall Assessment: APPROVE. Accurate, well-attributed, quotes verbatim, the docs-versus-changelog conflict honestly reported, policy length limits met, allowlist already covers github.blog, docs.github.com and github.com (no changes needed).