Repository navigation
π Activity Monitor: Ready for human merge offers Approve & merge / let hive merge; every needs-human state has its action - #15087
Conversation
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
|
@clubanderson is auto-promoted for contributor-lane merging on this PR: the last 5 merged PRs by this author in this hive's repositories have no revert or sentinel-alert evidence. CI, reviewer verdict, diff fences, soak, holds and branch protection still apply. |
|
@clubanderson this PR is held at Why:
Next: the hive does not push to your branch. Push a new head (or reply here); CI and the review swarm re-run automatically and the PR returns to the normal path. |
There was a problem hiding this comment.
Style / conventions review (read at head 38b57b5)
1. The merge-now note posted to the PR asserts an "untrusted author" cause for every 090 reason β src/pkg/dashboard/api_repo_item_decision.go:467 hardcodes "β¦parked this pull request for a human because its author is not a trusted merger on this repository", and the SPA confirm at src/pkg/dashboard/static/index.html:17583 says the same. But decisionForHumanMerge (stage_tracker_decision.go) is reached for any PR at 090, and pkg/stageengine/prstage/reasons.go:84-110 lists ten other HumanMergeReason* codes (workflow-files, tests-removed, branch-protection, manual-merge-policy, β¦). For those PRs the permanent GitHub comment would state a false reason β the opposite of the plain-words goal in #15078. Keep the comment reason-neutral (or thread human_merge_reason_text through):
body := fmt.Sprintf("Merged as maintainer by @%s from the Hive dashboard: the merge gate had parked this pull request at Ready for human merge, so the hive would not merge it unattended.", strings.TrimSpace(user))
and in index.html:17583 drop the clause because its author is not a trusted merger on this repo (or substitute items[0].human_merge_reason_text).
2. repoItemMergeDecisionAllowed (api_repo_item_decision.go:392-398) re-implements requireMergerOrOwnerRole (api_agent_control.go:494), which is the package's existing bar with 8 callers β including the very queue-automerge handler the new comment cites (api.go:1346). Reuse it so the two bars cannot drift:
- if !repoItemMergeDecisionAllowed(r) {
- jsonError(w, "merger or owner access required to merge or queue a pull request", http.StatusForbidden)
+ if !requireMergerOrOwnerRole(w, r) {
return
}and delete repoItemMergeDecisionAllowed (adjust the handler test's expected 403 message if it asserts on it).
β hive: agent=adjudicator-4 backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Security perspective (read at head 38b57b5; merge-base b26b481).
1. merge-now / queue-merge skip the "repository is managed by this hive" bar that the sibling queue-automerge endpoint enforces β src/pkg/dashboard/api_repo_item_decision.go:371-383 dispatches to handlePullRequestMergeDecision (:396) after only canToggleRepoHold (api_repo_hold.go:74-84, which short-circuits to allow for any verified owner) and the role check (:390-396). handleQueuePRAutoMerge (src/pkg/dashboard/api.go:1369-1372) additionally refuses with prQueueRepoAllowed(repo). Since the route takes {owner}/{repo} from the URL, a verified owner/merger can make the hive App merge (MergePRVia, :449) or approve+label any PR in any repo the App is installed on, not just project.repos. The other decisions in this handler already have that gap, but they comment/close β merge-now lands a commit on a base branch under the App's identity, so the new action should carry the same bar:
case repoItemDecisionMergeNow, repoItemDecisionQueueMerge:
if itemType != "pr" {
jsonError(w, decision+" applies to pull requests", http.StatusBadRequest)
return
}
if !repoItemMergeDecisionAllowed(r) {
jsonError(w, "merger or owner access required to merge or queue a pull request", http.StatusForbidden)
return
}
if !s.prQueueRepoAllowed(repoFull) {
jsonError(w, "repository is not managed by this hive", http.StatusForbidden)
return
}
s.handlePullRequestMergeDecision(ctx, w, repoFull, number, user, decision, comment)
2. merge-now enforces hold / sentinel only in the UI, not in the handler. stage_tracker_decision.go:254-256 disables the merge options with Release the hold first when held, and :503/:537 tells the operator the hive won't merge while the sentinel label is on β but handlePullRequestMergeDecision (api_repo_item_decision.go:396-468) checks only state/merged/self-author and then calls MergePRVia with no label read. queue-merge is covered because QueuePRAutoMergeWithNote refuses the sentinel label (src/pkg/github/client.go:2964-2969) and the sweep honours holds; merge-now has no equivalent, so a direct POST β¦/decision {"decision":"merge-now"} merges a held or sentinel-flagged PR from an untrusted author in one call. Suggest reading the labels before the merge branch:
base := state.BaseRef
if base == "" {
base = "the base branch"
}
+ if decision == repoItemDecisionMergeNow {
+ labels, err := gh.IssueLabels(ctx, repoFull, number)
+ if err != nil {
+ jsonError(w, "reading the pull request labels: "+err.Error(), http.StatusBadGateway)
+ return
+ }
+ if gh.IsHeldLabels(labels) {
+ jsonError(w, "pull request is on hold; release the hold first", http.StatusConflict)
+ return
+ }
+ if sl := s.stageTrackerSentinelLabel(); sl != "" && slices.ContainsFunc(labels, func(l string) bool { return strings.EqualFold(l, sl) }) {
+ jsonError(w, "the security check has an open alert on this pull request; remove `"+sl+"` first", http.StatusConflict)
+ return
+ }
+ }(use whatever accessor the stage tracker already uses to pass sentinelLabel into stageTrackerDecisionNeeded, stage_tracker_decision.go:141.)
Escaping of the new consequence/hint strings in static/index.html goes through esc() throughout; no injection there. Audit records and expectSHA pinning (:449) look right.
β hive: agent=adjudicator-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
docs-currency β read at head 38b57b5. Two operator-facing text gaps; the rest of the docs match the code (options per state in stage_tracker_decision.go, tooltip text in stage_tracker.go, toast/badge words in index.html).
- Merge note on the PR states the wrong reason for non-trust 090 parks.
src/pkg/dashboard/api_repo_item_decision.go:467hardcodes "β¦because its author is not a trusted merger on this repository", and the SPA confirm atsrc/pkg/dashboard/static/index.html:17583says the same. ButdecisionForHumanMerge(stage_tracker_decision.go:216-220) and your own docs row (dashboard-sections.md:717) cover 090 for workflow files, manual-merge path and branch protection too β on those PRs the public comment left on the contributor's repo will assert a falsehood about the author. Make the sentence reason-neutral (the row'sreason/human_merge_reasontext is already in the payload if you want to name it):
body := fmt.Sprintf("Merged as maintainer by @%s from the Hive dashboard: the merge gate had parked this pull request for a human to merge.", strings.TrimSpace(user))
and in index.html:17583 drop "because its author is not a trusted merger on this repo" in favour of "The merge gate parked it for a human; this is you taking that call."
- Audit key doc does not match what is recorded.
src/docs/dashboard-sections.md:717says "the audit log recordsrepo_item_decision_merge,source: dashboard". The dashboard record atapi_repo_item_decision.go:458carriesrepo/number/type/head/author/base/shaβ nosource.dashboardis thepathof the separatepr_mergedrecord written byRecordPRMergedAudit(src/pkg/github/client.go:4487, viaMergePRVia).operator-reference.md:472already has it right ("audit pathdashboard"); align the sections doc:
| PR at 090 *Ready for human merge* (untrusted author, workflow files, manual-merge path, branch protectionβ¦) | **Approve & merge as maintainer** (merges #N into *base* now via hive's identity, attributed to you β a comment on the PR names you; the audit log records `repo_item_decision_merge` and the App's `pr_merged` with `path: dashboard`) Β· **Approve β let hive merge** (records your approval as the owner decision β an APPROVE review and the automerge label; the hive's merge sweep merges it on its next pass once CI is green and the head has soaked) Β· Give direction (hive-authored PRs only) Β· Close as superseded Β· Close as won't fix Β· Hold. Both merge actions need the merger or owner role and confirm first (with *Don't ask again*); the toast reads *merged* / *queued for merge*; the badge records `merge-now` / `queue-merge`. |
Minor, same files: dashboard-sections.md:713 dates the quote 2026-10-29 while #15078 says 2026-10-28; operator-reference.md:473 "an APPROVED review in your name" β the review is the App's, with Approved by @you in its body (client.go:2974-2993); "attributed to you" is the accurate phrase.
β hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Correctness (read at 38b57b5)
1. The 090 decision is reason-agnostic, but the two lead options only hold for untrusted-author β src/pkg/dashboard/stage_tracker_decision.go:208-258 (decisionForHumanMerge) enables Approve & merge as maintainer and Approve β let hive merge for every HumanMergeReason code (stage_tracker.go:1261-1271: workflow-files, tests-removed, manual-merge-policy, branch-protection, merge-method, auto-merge-off, unauthorized, reviewer-deferred), with the same consequence text.
queue-merge: the sweep's queued lane re-applies the carried fence andhumanMergePathGate(src/pkg/github/automerge/automerge_sweep.go:2545-2552, "a merger queuing the PR is not the same as a person merging it"), so for workflow-files / tests-removed / manual-merge-policy the promise "the hive's merge sweep merges #N on its next pass once CI is green" is false β the PR stays parked and the operator's approval goes nowhere, with no feedback in the row.merge-nowonworkflow-files/unauthorized: the App lacks the permission (that is why the row is at 090), soMergePRVia(pullrequest.go:404) fails and the handler answers 502 (api_repo_item_decision.go:453-456). The option should be disabled with the reason, not offered and failed.merge-nowontests-removed/manual-merge-policy:MergePRViaonly runsCheckIrreversible(pullrequest.go:415), so the dashboard now merges through the hive's identity exactly the PRs those policies say the hive never merges. That is a deliberate operator call, but the comment it posts (api_repo_item_decision.go:470) hard-codes "because its author is not a trusted merger", which is wrong for these reasons.
Suggested fix β gate by the reason code the function already derives:
func decisionForHumanMerge(it StageTrackerItem, held bool) *StageTrackerDecision {
base, approver := stagelabels.HumanApprovedReason(it.Reason)
base = strings.TrimSpace(base)
- _, word, _ := stageTrackerHumanMergeReason(it.Reason)
+ code, word, _ := stageTrackerHumanMergeReason(it.Reason)
@@
- mergeNow := decisionOption(decisionActionMergeNow, "Approve & merge as maintainer", !held, "Release the hold first")
+ mergeWhy, queueWhy := "", ""
+ switch code {
+ case stageengineprstage.HumanMergeReasonWorkflowFiles, stageengineprstage.HumanMergeReasonUnauthorized:
+ mergeWhy = "The hive's GitHub App is not permitted to merge this; click Merge on GitHub"
+ queueWhy = mergeWhy
+ case stageengineprstage.HumanMergeReasonTestsRemoved, stageengineprstage.HumanMergeReasonManualMergePolicy,
+ stageengineprstage.HumanMergeReasonAutoMergeOff, stageengineprstage.HumanMergeReasonMergeMethod,
+ stageengineprstage.HumanMergeReasonBranchProtection:
+ queueWhy = "The merge sweep applies the same rule that parked this PR, so queuing it changes nothing; merge now or on GitHub"
+ }
+ if held {
+ mergeWhy, queueWhy = "Release the hold first", "Release the hold first"
+ }
+ mergeNow := decisionOption(decisionActionMergeNow, "Approve & merge as maintainer", mergeWhy == "", mergeWhy)
mergeNow.Consequence = "merges " + ref + " into " + into + " now via hive's identity, attributed to you"
- queue := decisionOption(decisionActionQueueMerge, "Approve β let hive merge", !held, "Release the hold first")
+ queue := decisionOption(decisionActionQueueMerge, "Approve β let hive merge", queueWhy == "", queueWhy)and pass word/reason into renderRepoItemDecisionMergeComment instead of the fixed "not a trusted merger" sentence. Add the non-untrusted-author codes to TestStageTrackerNeedsHumanStatesCarryAnAction (today every 090 fixture uses ContributorLaneReasonAuthorAssociation, stage_tracker_decision_test.go:340-342, so the table never exercises the other reasons).
Advisory: tide reports the branch conflicts with v5; needs a rebase before it can land.
β hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Intent alignment vs #15078 β the 090 path (merge-now / queue-merge) is wired end to end, but two of the newly-added "every needs-human state carries an action" states render a lead action that does nothing when clicked.
Lead decide:* options are dropped at dispatch for rows that are not "parked" (071/072 human PR, needs-human outside 045/085/090).
stageTrackerLeadActions builds the menu from decision_needed.options via stageTrackerDecisionOf (static/index.html:16762-16775), but both click paths β menu (index.html:16953 β stageTrackerDoAction) and popover (index.html:17147 β stageTrackerDecisionRun β runRowAction) β land in runRowAction's decide: branch (index.html:17815-17820), which maps each row through stageTrackerDecisionTarget and drops nulls. stageTrackerDecisionTarget only recognises 045/085, the needs-decision pill, superseded, manual_merge or human_merge_reason (index.html:16673-16693); your own test pins stageTrackerDecisionTarget({stage: 70, modifiers: []}) === null (stage_tracker_row_actions_test.go:239). So:
- the
waitingfixture (human PR at 072,stage_tracker_row_actions_test.go:1050-1057) renders Close as superseded / Close as won't fix at the top, but clicking either returns{ toasts: [], pending: [] }silently β no request, no toast; - the new Resolve β lift needs-human (
decide:approve,stage_tracker_decision.go:646-653) on a needs-human row at any stage other than 045/085/090 does the same.
The server side would accept both (RecordOperatorIssueDecisiontreatsneeds-humanas parked,issue_unpark_command.go:749-759;handlePullRequestDecisionpassesLiftNeedsHuman: true,api_repo_item_decision.go:506-509), so this is purely the SPA target resolution.TestStageTrackerRowActionsLeadWithDecisionOptionsonly checks rendering; norunRowActioncase covers these states.
Fix: fall back to the asking item when the row itself (or its one asking nested PR) carries the option:
} else if (id.indexOf('decide:') === 0) {
// A row decides its parked PR when it has one (the issue's stage is
// only mirrored from it); dedupe when the PR row is selected too.
// A row that is not parked but whose decision_needed names the
// action (071/072 human PR, needs-human at any stage) decides itself
// (hivecommons/hive#15078).
const seen = {};
items = items.map(it => {
const t = stageTrackerDecisionTarget(it);
if (t) return t;
const dec = typeof stageTrackerDecisionOf === 'function' ? stageTrackerDecisionOf(it) : null;
const opts = dec && dec.item && dec.item.decision_needed && Array.isArray(dec.item.decision_needed.options) ? dec.item.decision_needed.options : [];
return opts.some(o => o && o.action === id) ? dec.item : null;
}).filter(t => t && !seen[stageTrackerItemKey(t)] && (seen[stageTrackerItemKey(t)] = true));
if (!items.length) return { toasts: [], pending: [] };
}
and add a runRowAction('decide:close-wontfix', waiting) / runRowAction('decide:approve', <needs-human at 060>) case asserting a request is issued, so the renderer test and the dispatch agree.
β hive: agent=adjudicator-3 backend=copilot model=claude-fable-5.1 copilot=1.0.88
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
β¦ed conflicting PRs; dismiss self-inflicted changes_requested (#15104) * π Activity Monitor: Rebase, then merge as maintainer; autonomous rebase-resolve with loop guard; approve on a dirty PR queues the rebase; dismiss self-inflicted changes_requested Squashed and rebased onto v5 after #15087 and #15074 landed. Fixes #15103 Signed-off-by: Andrew Anderson <andy@clubanderson.com> * test: fix fixtures after the v5 rebase (mock fields, failingEntry rebase fields, import ratchet, window shim) Signed-off-by: Andrew Anderson <andy@clubanderson.com> * test: lift pkg/rebasemerge coverage above its floor; drop an ineffectual assignment Signed-off-by: Andrew Anderson <andy@clubanderson.com> * test: compare the shared singletons through named values (staticcheck SA4000) Signed-off-by: Andrew Anderson <andy@clubanderson.com> * security: rebase-then-merge fails closed on a head the hive did not produce; document the new API shapes The approval is bound to the head the operator read. A contributor PR whose head moves after the click now counts as rebased only when GitHub reports the hive (App bot or pane identity) as the commit's author or committer; any other head settles the follow-through as failed (foreign-head) with nothing merged, a note on the PR, and 'Approve the new head & merge' as the way out. api-reference.md gains the 202 rebase-queued answer, the rebase response field, and the rebase_merge / rebase row fields. Signed-off-by: Andrew Anderson <andy@clubanderson.com> * merge train: hand a dirty member to the fixer's rebase-resolve episode; the exhausted row's rebase-then-merge option goes live v5 landed the conflict-aware merge train (#15157) with a RebaseResolver hook reserved for this PR and a duplicate rebaseMergeFollowThroughLabel. The hook now opens the shared rebase episode (requested_by: merge-train); the mergeNowWired flag and its disabled reason go, since the option it guarded is wired here. Signed-off-by: Andrew Anderson <andy@clubanderson.com> * test: approve toast wording follows #15213 Signed-off-by: Andrew Anderson <andy@clubanderson.com> * rebase onto v5: drop duplicate nonEmptyString, lift STAGE_TRACKER_ICONS for the history badge in the rebase-then-merge row test Signed-off-by: Andrew Anderson <andy@clubanderson.com> * writeback ratchet: baseline the two operator-solicited comments of the rebase-then-merge follow-through Signed-off-by: Andrew Anderson <andy@clubanderson.com> * automerge: superseded question precedes the label block (#15278) while a needs-rebase label on a merely-behind head stays stale Signed-off-by: Andrew Anderson <andy@clubanderson.com> --------- Signed-off-by: Andrew Anderson <andy@clubanderson.com>
β¦ un-park/re-park loop, name the missing App permission (#15134) * Model capability gaps as first-class blockers so delegated decisions stop re-parking loops An issue the scanner parks because the GitHub App cannot push .github/workflows/* (no workflows permission) is a capability gap, not a question: a direction decision cannot clear it, and "Let hive decide" un-parked #14976 only for the sweep to re-apply the same needs-human text an hour later. - pkg/blocker: structured blocker kinds, the <!-- hive:blocker kind=β¦ --> marker, prose inference, and a persisted ledger with a delegated-unpark loop guard and a cleared-at staleness guard. - Record blockers where comments are read: issue_comment webhook, task-list sweep (adds the marker to the remainder comment), gh-wrapper (adds the marker to a needs-human comment that names a missing capability). - Stage engine: blocker parks 045 with "blocked: <short>"; needs-human beside approved-direction keeps 045 and clears the ack (needs-human is the newer signal by construction). - Dashboard: decision_needed.blocker/options populated with real handlers (App settings link, Re-check permissions, close as partially addressed, wait; patch hand-off disabled with its reason), delegated decision short-circuits to no-effect with the glossary 'blocker' reason, the row pill reads "Blocked Β· App lacks workflows permission", β β Security and /api/status.capabilities name missing permissions and blocked counts. - Tests: Go (blocker, ledger, webhook, sweep marker, stage facts/derive, dashboard handlers, glossary) and node fixtures (pill, options, popover, capability text); gh-wrapper bash tests. - Docs: labels-and-control-signals (blocker marker + kinds + reconcile rule), dashboard-sections, operator-reference (App permissions table), api-reference; changelog fragment. Fixes #15099 Signed-off-by: Andrew Anderson <andy@clubanderson.com> * Blocker options carry a plain-words consequence for the popover Follows #15087's option consequences so a blocked row's ways out say what each one does. Signed-off-by: Andrew Anderson <andy@clubanderson.com> * docs: stable anchor for the Blockers section Signed-off-by: Andrew Anderson <andy@clubanderson.com> * fix: close-partial response carries the result by pointer Signed-off-by: Andrew Anderson <andy@clubanderson.com> * fix: pkg/github blocker import alias, ledger loop-guard timing, test call sites, import ratchet - pkg/github already has a test helper named blocker; import pkg/blocker as itemblocker there. - Ledger.Record drove the loop guard off the preserved first-report time, so a capability re-report after a delegated un-park never tripped it; use the observation time. Observe returns changed=false for a report older than the entry's cleared-at. - stageTrackerDecisionNeededWith test calls pass the clock. - Allow pkg/dashboard β pkg/blocker in the import ratchet (the blocker ledger is the seam the tracker, delegate and status read). Signed-off-by: Andrew Anderson <andy@clubanderson.com> * fix: keep the import allowlist sorted Signed-off-by: Andrew Anderson <andy@clubanderson.com> * Fix recheck/held tests, document blocker routes in openapi.json, drop duplicate JS const Signed-off-by: Andrew Anderson <andy@clubanderson.com> * test: cover blocker kind words, ReasonSentence caps and ledger edge cases; resync api-reference citations after rebase Signed-off-by: Andrew Anderson <andy@clubanderson.com> * fix: mark the close-partial blocker writes as operator-solicited for the write-back ratchet Signed-off-by: Andrew Anderson <andy@clubanderson.com> * docs: resync api-reference citations after rebase onto v5 Signed-off-by: Andrew Anderson <andy@clubanderson.com> --------- Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Fixes #15078
Observed
hivecommons/hotshot#316 β nested PR hotshot#317 sat in the Merge gate as Ready for human merge Β· untrusted author π€ (
stage/090-ready-for-human-merge). The issue row's β― menu offered only Hold / Re-evaluate now / Mark needs-humanβ¦ / Open on GitHub / Copy link β "this claims it needs human merge but there is no option to take that action".Why #317 is "untrusted author" (verified from hive's hive config + live tracker payload)
The PR author is @Karibhou (a human,
author/clanker, registry tiernewcomer), not the hive App. The row's reason iscontributor-lane:author-associationβhuman_merge_reason: untrusted-author: Karibhou has no OWNER/MEMBER/COLLABORATOR association on hotshot and notrusted/mergertier (auto_merge.contributor_lane.require_tiersdefault[trusted, merger]), and is not underdashboard.authorized_users, so neither the contributor lane nor the trusted-authors lane vouches for the PR. Fix: π₯ Contributors tier select, GitHub collaborator, or Authorized users withmergerrole (β Security). The pill tooltip now says exactly this.Changes
Backend (Go)
decision_neededfor a PR at 090 leads with Approve & merge as maintainer (decide:merge-now, consequence merges #N into main now via hive's identity, attributed to you) and Approve β let hive merge (decide:queue-merge, records your approval; the merge sweep merges on its next pass once CI is green and the head has soaked), then Give direction / Close as superseded / Close as won't fix / Hold. NewConsequencefield on options.POST /api/repos/{o}/{r}/items/{n}/decisionacceptsmerge-nowandqueue-merge(PRs only; verified owner ormergerrole; refuses merged/closed PRs and your own PR). merge-now comments then merges viaMergePRVia(..., PRAuditPathDashboard); queue-merge usesQueuePRAutoMergeWithNote. Audited asrepo_item_decision_merge/repo_item_decision_queue_merge; ledgerlast_operator_action.actionmerge-now/queue-merge; resultmerged/queued-for-merge.stageTrackerNeedsHumanstate:needs-humanlabel β Resolve β lift needs-human;hive:unauthorizedβ Re-evaluate now; human-authored PR at 071/072 β close options; any other flagged stage β parked fallback (Re-evaluate / Hold / Mark needs-human). Hold set by operator keeps Release hold leading with the state's own options disabled with Release the hold first. 085 / 045 / 015 / reporter-trust hold unchanged (built on π Activity Monitor: PR-scoped row actions on nested PRs; Let hive decide per PR, never silently oneΒ #15056, β¨ Activity monitor: decision badge shows what you chose and what hive did nextΒ #14992, π fix(stage): a contributor PR with no linked issue is not waiting on the hiveΒ #15020's work).TestStageTrackerNeedsHumanStatesCarryAnAction: every stage/modifier the tracker flags needs-human yields non-empty options with an enabled action or a plain-wordsreason_if_disabled. PlusTestStageTrackerDecisionForHumanMerge, handler tests for merge-now / queue-merge / guards, operator-action mapping cases.stageTrackerUntrustedAuthorWhy): why + config key + where in β/π₯.SPA (
dashboardSetIntervalonly,[hidden], style ratchet respected)stageTrackerRowActions) leads with the server'sdecide:*options above a separator, consequence as hint; works for issue rows with one nested PR (targets the PR) and PR sub-rows viarunRowAction(action, row).TestStageTrackerRowActionsLeadWithDecisionOptions(node fixture via stdin).Docs:
dashboard-sections.md(row actions per state table + untrusted-author wording),operator-reference.md(merge-now / queue-merge table),changelog.d/fixed-ready-for-human-merge-actions.md.What hotshot#317's issue row β― menu now shows
Approve & merge as maintainer Β· Approve β let hive merge Β· β Β· Decide PR #317β¦ (Let hive decide / Approve / Close as superseded byβ¦ / Won't fixβ¦ / Give direction β disabled: human author) Β· Hold Β· Re-evaluate now Β· Mark needs-humanβ¦ Β· Open on GitHub Β· Copy link