Repository navigation
Description
What happened?
while setting up the repo i noticed some changelog fragments are still under src/changelog.d/ instead of changelog.d/ at the repo root. compile-changelog.sh only reads the root folder, so these entries never got into CHANGELOG.md.
on v5 there are 14 of them, and none of their PRs show up in CHANGELOG.md:
- added-activity-monitor-accept-triage.md (✨ dashboard: Accept triage in the activity monitor row menu #13805)
- added-live-run-line.md (✨ Live per-run cost · tokens · elapsed · tests · files line on agent cards #14580)
- added-merge-issue-author-credit.md (✨ Credit issue authors on clanker-opened PRs at merge time #14585)
- added-tracker-sentinel-lane-text.md (✨ dashboard: activity monitor names the sentinel alert in the merge-gate lane #13760)
- changed-crux-cmd-hive-runtime-seam.md (🌱 cmd/hive: consume the agent runtime through the crux.Runtime seam #14220)
- changed-crux-runtime-seam.md (🌱 crux: Runtime seam for the sidecar plan, stage 1 #14208)
- changed-reviewer-l3-l4-packs.md (✨ acmm: reviewer agent joins the L3 and L4 packs #13609)
- fixed-13817-watchdog-producing-cadence.md (🐛 watchdog: widen the no-production window to the agent's own cadence #13821)
- fixed-13818-ci-gate-held-twin-runs.md (🐛 ci gate: ignore held runs whose workflow already passed on the head and name same-repo bot PRs #13822)
- fixed-adopted-pr-rebase.md (🐛 Adoption: a hive-adopted PR that conflicts with its base gets rebased by hive #13280)
- fixed-contributor-claim-starvation.md (🐛 claims: contributor disconnects release the issue and flapping no longer starves hive agents #13758)
- fixed-cost-per-agent-copilot-sessions.md (🐛 cost: scan per-agent Copilot session trees and split cached input #13486)
- fixed-supersession-sweep-stage-machine.md (🐛 Route supersession sweep escalation through the stage engine and let the hive resolve its own superseded PRs #13807)
- fixed-upgrade-drain-visible-to-hub.md (🐛 Hub holds upgrades the spoke is draining instead of orphaning them #14322)
on v6 there are 81, and 77 of them are not in v6's CHANGELOG.md.
there is also a leftover changed-gate-fold-tap-target.md at the repo root from #17515. its entry is already in CHANGELOG.md under v5.211.11, so that file can just be removed.
Expected behavior
every merged fragment ends up in CHANGELOG.md, and nothing is left in src/changelog.d/ except its README.
Reproduction steps
- git fetch origin
- git ls-tree -r origin/v5 --name-only | grep '^src/changelog.d/[a-z]*-'
- pick any PR number from the list above and search CHANGELOG.md for it. it is not there.
Environment
- Branch/version: v5 at 22708bb, v6 at 15ad6de7fc
- Deployment method: not deployment related, repo contents only
- Operating system or Kubernetes version: n/a
- Relevant configuration (redact secrets): n/a
Logs or screenshots
on #14580, #13805 and #14322 the changelog-fragment-guard check shows FAILURE, followed by several SKIPPED runs from label events.
Additional context
all of these were added between oct 20 and oct 27. the guard did fail on those PRs, but github counted the later skipped label runs as passing. #14767 fixed that on oct 28, so new ones should be blocked now and only the cleanup is left. the earlier cleanups (#11928, #12675) ran before these were added.
one question before i send a PR: for releases that already shipped, should i move the fragments into changelog.d/ so the next release picks them up, or add them to CHANGELOG.md under the versions where they actually went out? i can do v5 first and v6 in a separate PR, or whatever you prefer.
Activity
github-actions commentedon Nov 15, 2026
Welcome to the Hive Commons community!
Thanks for opening your first issue here! Every issue helps make the project better.
To work on this issue, comment
/assignto assign yourself (use/unassignto remove).Find issues to work on: look for the
good first issueandhelp wantedlabels.Hope you have a great time here!
For Maintainers:
/good-first-issue|/help-wantedItx-Psycho0 commentedon Nov 15, 2026
/assign
hivecommons-hive commentedon Nov 15, 2026
Thanks — this hive only works issues from OWNER, MEMBER, COLLABORATOR automatically. A maintainer can admit this one by adding the label
triage/accepted(configured inissue_filter.reporter_trust.untrusted_require_labels). Until then the hive will not claim, label, or open PRs for it. See https://github.com/hivecommons/hive/blob/HEAD/GOVERNANCE.md#reporter-trust-and-escalation for what these terms mean and how to become a trusted reporter.clubanderson commentedon Nov 16, 2026
Triage: confirmed. Thanks for the careful inventory, your counts are exact.
Evidence (v5 at 42bc2ba, v6 at 97b0461)
src/scripts/compile-changelog.shandtagged-release.ymlcompile and delete only the repo-rootchangelog.d/.src/changelog.d/holds README plus 14 fragments. None of their text and none of their PR numbers (🐛 Adoption: a hive-adopted PR that conflicts with its base gets rebased by hive #13280, 🐛 cost: scan per-agent Copilot session trees and split cached input #13486, ✨ acmm: reviewer agent joins the L3 and L4 packs #13609, 🐛 claims: contributor disconnects release the issue and flapping no longer starves hive agents #13758, ✨ dashboard: activity monitor names the sentinel alert in the merge-gate lane #13760, ✨ dashboard: Accept triage in the activity monitor row menu #13805, 🐛 Route supersession sweep escalation through the stage engine and let the hive resolve its own superseded PRs #13807, 🐛 watchdog: widen the no-production window to the agent's own cadence #13821, 🐛 ci gate: ignore held runs whose workflow already passed on the head and name same-repo bot PRs #13822, 🌱 crux: Runtime seam for the sidecar plan, stage 1 #14208, 🌱 cmd/hive: consume the agent runtime through the crux.Runtime seam #14220, 🐛 Hub holds upgrades the spoke is draining instead of orphaning them #14322, ✨ Live per-run cost · tokens · elapsed · tests · files line on agent cards #14580, ✨ Credit issue authors on clanker-opened PRs at merge time #14585) appear anywhere in v5's CHANGELOG.md. So those entries shipped in the late-October v5 releases with no release notes.src/changelog.d/holds 81 fragments: the same 13 v5 files (v6 had already moved 🐛 Adoption: a hive-adopted PR that conflicts with its base gets rebased by hive #13280's) plus 68 that exist only on v6. Of those 68, 4 are already recorded in v6's CHANGELOG.md (🌱 admin MCP: reads for the agent, run, health, prompt-history and config data the dashboard already serves #11943 reworded, 🌱 admin MCP: start records, crash traces and out-of-memory evidence without host access #11945, 🐛 [v6] Conflicting Renovate PR withrebaseWhen: nevernever getsneeds-human, so nothing shows a person must act #11558 under v6.0.0; 🐛 [v6] /api/nous/status: baseline_pct is unbounded (reports 1565%) once snapshots pass NousBaselineTarget #10026 under v5.114.0), which leaves 77 dropped, as you found.changelog-fragment-guardalready rejectssrc/changelog.d/(🐛 fix(changelog): reject fragments under src/changelog.d/ and point authors to the repo root #11926), and it did fail on these PRs. The skipped label-event runs after it counted as green, and 🐛 ci: never report the fragment guard as skipped on label events #14767 closed that on 10-28. Also, at the timeAGENTS.md(whichCLAUDE.mdlinks to) said onlychangelog.d/<...>.md, without "repository root", which is ambiguous from inside thesrc/Go module. 🌱 AGENTS.md: keep tone and workflow rules; move repository facts into scoped knowledge entries #17051 added the root wording on 11-07.src/changelog.d/on v6 is 10-27.changed-gate-fold-tap-target.mdis already gone from v5.On your question: the repo precedent (#11928, #12675) is to move strays into the root
changelog.d/so the next release carries them, not to backfill sections that already shipped. Both PRs below follow that.Fix
-prefix).compile-changelog.shnow fails loudly while any fragment sits insrc/changelog.d/, so a stray that slips past the PR guard turns the next release red instead of disappearing. Guard tests and a repo invariant are added totest-compile-changelog.sh.AGENTS.md, both READMEs and the knowledge entry now spell out../changelog.d/for anyone working fromsrc/.changelog.d/fixed-adopted-pr-rebase.mdon v6, because v6.0.0 already carries that entry.Sorry for jumping ahead of your offered PR; this one was tied to release tooling. Reviews on either PR are very welcome.
🐛 changelog: recover v6-only fragments stranded in src/changelog.d
hivecommons-hive commentedon Nov 16, 2026
Fixed by #18361 (merged 0cd0a4f). Closing. Reply /reopen if the problem persists.
github-actions commentedon Nov 16, 2026
@Itx-Psycho0 this issue was closed by @hivecommons-hive[bot] as completed.
If this was closed by a merged PR, the fix only takes effect once it ships in a tagged release and your deployment updates to it — check your running version against the fix before concluding the fix failed.
If you are still seeing the problem, comment
/reopenand it will be reopened for you — you do not need write access. Please add what you are still observing (running version, logs, steps) so it is not closed the same way again.🐛 changelog: recover 14 fragments stranded in src/changelog.d and fai…
clubanderson commentedon Nov 16, 2026
Fixed by #18360 (v5, merged) and #18361 (v6, merged). v5's 14 strays and v6's 64 v6-only strays now sit in the root changelog.d/, so the next releases carry them. 4 v6 strays were dropped because their entries were already recorded. The remaining 13 shared files reach v6 through the next top-up, which should drop changelog.d/fixed-adopted-pr-rebase.md on v6 (already in v6.0.0). From now on, compile-changelog.sh fails the release while any fragment sits under src/changelog.d/.