Skip to content

πŸ› fix(contribute): a needs-decision park posts its question - #16837

Merged
hivecommons-hive[bot] merged 2 commits into
hivecommons:v6from
Danathar:fix/park-says-why
Nov 6, 2026
Merged

hivecommons-hive[bot] merged 2 commits into
hivecommons:v6from
Danathar:fix/park-says-why

Conversation

@Danathar

@Danathar Danathar commented Nov 5, 2026 •

Copy link
Copy Markdown
Contributor

Executive summary

When a hosted contributor's agent ends a task with a needs_decision verdict, the issue goes to stage/045-needs-decision. The contributor relay applies that label itself, and the agent is only asked, in the task prompt, to leave the question as a comment. Nothing checked that it did. projectbluefin/common#1605 (j1mc's agent, 2026-11-01 05:58Z) and projectbluefin/utah#1036 (Karibhou's agent, 15:36Z) were parked that way with no comment. Their only explanation was the dashboard's "Parked until a maintainer decides".

The hub already receives the verdict's reason. With this PR, when nobody has commented on the issue since the task was assigned, the hub posts that reason as the question once, and the recommendation when the reason names one.

What

Situation (standing needs-decision verdict on a GitHub issue) Before After
Agent left no comment 045, no question Hub posts Question: with the reported reason
Reason names a recommendation (…; recommended: B, because …) β€” Also Recommendation: B, because …, so /hive approve accepts it
Reason names none β€” Says "named no preferred option" and points at /hive decision; never says "recommend"
The reporting contributor explained the park, or the hub already posted for this task β€” Nothing posted
Only hive-marked notices since assignment (utah#1036's good-first-issue withdrawal) β€” Not an explanation: the question is posted
Repo-local gate park the hub releases (#13059) Released Released (no park comment)

The task prompt now asks for HIVE_VERDICT: no_work_needed β€” decision: <the question>; recommended: <the option, because why>.

Why

  • The two parks come from the hive log at those times. Both were [contribute-ws] task complete … verdict=needs_decision … needs_decision_label=stage/045-needs-decision. Neither issue got a comment from the agent or the hub.
  • markIssueNeedsDecision in bin/contributor-relay.js applies the label. The hub only recorded the reason in its ledger.
  • The un-park menu treats any hive comment that mentions "recommend" as stating a recommendation. A hub note saying "no recommendation" would therefore make /hive approve offer an empty choice, so that word is avoided.

How

  • contribute_needs_decision_park.go (new):
    • postNeedsDecisionPark runs detached after a needs-decision verdict that stands, on GitHub-backed issue tasks only.
    • It lists comments since the task's assignment time, with a minute of slack. A non-trivial comment by the reporting contributor that isn't a hive-marked notice counts as the agent's explanation, as does an earlier hub park comment. A maintainer's or bystander's remark does not. A wrong-base fallback park keeps its own note and gets no question. If neither exists, it posts one marked comment (<!-- hive:contribute-park task=… -->).
    • The reason is sanitized as everywhere else: no mentions, no HTML-comment delimiters, capped.
  • contribute_ws.go: wires the call after the gate-park reclassification, so a released gate park is never commented on. It also adds a test seam.
  • contribute_task_prompt.go, docs/contributor-relay.md: the decision form now asks for the question and a recommendation.
  • notification_write_sites.go, docs/notifications.md: the new writer is inventoried as attention.

Tests

Test Covers
TestNeedsDecisionParkComment marker, question, recommendation carried (incl. "3.14"), no "recommend" when none, empty reason, sanitizing
TestParkCommentExplains reporter prose vs a maintainer's or bystander's remark vs hive-marked notices vs the hub's marker
TestNeedsDecisionParkPost fake GitHub: no comments / gfi notice only / older comment / agent explained / already posted; since bound
TestNeedsDecisionParkSkipsNonIssueTasks PR, Linear, malformed tasks, nil hub
TestNeedsDecisionParkWireShape real task_complete handler posts for #1605's verdict; a released gate park posts nothing
TestNeedsDecisionParkDue, TestNeedsDecisionParkNotForWrongBaseFallback gate park and wrong-base fallback get no hub question; the wire test no longer skips
Mutants 17/17 caught (10 + 7 on the review fixes)
go test dashboard, github, writeback pass
check-docs-citations.py src/docs 2006 verified

Notes

πŸ€– Generated with Claude Code

https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1

A hosted contributor's needs-decision verdict parks the issue at 045 through
the relay's label alone; the agent is only asked to comment.
projectbluefin/common#1605 and projectbluefin/utah#1036 (2026-11-01) sat at
045 with no question. When nobody commented after the task was assigned,
the hub now posts the reported question once (marker
`hive:contribute-park task=…`), with the recommendation when the reason
names one and without the word otherwise, so the un-park menu offers no
empty `/hive approve`. The task prompt asks for
`decision: <question>; recommended: <option, because why>`.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1
Signed-off-by: Danathar <Danathar@users.noreply.github.com>
@kubestellar-prow kubestellar-prow Bot added the dco-signoff: yes Indicates the PR's author has signed the DCO. label Nov 5, 2026
@kubestellar-prow

Copy link
Copy Markdown
Contributor

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign nicholasjackson for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@kubestellar-prow kubestellar-prow Bot added the size/L Denotes a PR that changes 100-499 lines, ignoring generated files. label Nov 5, 2026
@github-actions

github-actions Bot commented Nov 5, 2026 •

Copy link
Copy Markdown
Contributor

Open PR overlap check

Base branch: v6

Conflicts are checked with every PR merged onto the current tip of v6, so a PR that only needs a rebase is not reported as conflicting with its siblings.

No conflicting open PRs or same-file overlaps were found.

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Test review (quality lane) β€” read at head 317e891. The behaviour change ships failing-without-it tests: TestNeedsDecisionParkPost ("no comments" / "only the hive's own GFI notice") would post nothing on v6 today, and TestNeedsDecisionParkWireShape pins the handleTaskComplete seam. Two gaps, neither blocking:

  1. parkCommentExplains never looks at the comment's author. Any β‰₯40-char comment created after assignedAt suppresses the hub's question β€” including a maintainer's or a bystander's unrelated remark during the task window, which is exactly the case where the question is most needed. No table row exercises a third-party comment; consider adding one (user.login != reporter β†’ still post) and, if you agree, restricting the "explained it" match to the reporter's login (and the hub's own marker).
  2. The second half of TestNeedsDecisionParkWireShape ends in t.Skipf when the harness doesn't re-offer #77, so the "gate park is released, not posted" branch is only asserted when the harness happens to cooperate. A direct call to postNeedsDecisionPark with gateParkReason (like TestNeedsDecisionParkSkipsNonIssueTasks does for Kind: "pr") would make that assertion unconditional.

β€” hive: agent=quality backend=copilot model=claude-fable-5.1 copilot=1.0.88

@hivecommons-hive

hivecommons-hive Bot commented Nov 5, 2026 •

Copy link
Copy Markdown
Contributor

Hive status β€” updated in place; edits do not notify. Full history is in the Activity Monitor.

Review findings from reviewer-2 (folded here because this head already carries its hive review):

Docs-currency pass at 317e891 β€” advisory only, no blocker.

src/docs/contributor-relay.md:1626 still says no comment is posted for a decision park. The paragraph the PR edits (line 1624) now documents the hub's <!-- hive:contribute-park task=… --> comment, but two lines later the section still reads "The relay posts no extra comment: the contributor's assessment comment is the human-readable record." A reader of that paragraph gets both claims. The relay statement is still true; it just needs to name the hub fallback that postNeedsDecisionPark (src/pkg/dashboard/contribute_needs_decision_park.go:109) adds.

The hub books the issue for the full with-PR cooldown immediately and records a `needs_decision` ledger/run-log marker. The relay applies the label advertised by the hub in `auth_ok` (`hub.contribute_needs_decision_label`, default `needs-decision`) using the same task credential path as the `blocked` label. Empty config disables relay labelling, but the cooldown still applies. The relay posts no extra comment: the contributor's assessment comment is the human-readable record, and when there is none the hub's marked park comment above stands in for it. The label is never cleared automatically; removing it is the maintainer signal that the decision has been made. The configured label is also treated as a contributor skip label, so a custom label such as `2-discussing` keeps the issue out of the offer queue until a human removes it.

Checked and fine: the notifications.md row names needsDecisionParkPost, which is the function enclosing the CreateIssueCommentBudgeted call (contribute_needs_decision_park.go:150) and so matches how TestNotificationWriteSitesInventory attributes write sites; the decision: <…>; recommended: <…> form is updated at its only two sites (contribute_task_prompt.go:659, contributor-relay.md:1624); /hive decision is a real command (api_repo_item_decision.go:41).

Review findings from adjudicator-3 (folded here because this head already carries its hive review):

Correctness review (head 317e891)

1. A wrong-base verdict that falls back to needs-decision now gets two hub park comments. resolveWrongBaseVerdict returns completionVerdictNeedsDecision with park=true when the target branch is missing/unvalidatable (src/pkg/dashboard/contribute_wrong_base.go:46-51), and the handler launches parkWrongBaseIssue, which posts a <!-- hive:wrong-base-park --> comment (contribute_ws.go:4488-4496, contribute_wrong_base.go:111-118). Because verdict is still needs_decision afterwards, the new block at contribute_ws.go:4542-4545 also launches postNeedsDecisionPark. parkCommentExplains treats any <!-- hive: comment that is not its own marker as not an explanation (contribute_needs_decision_park.go:52-60), so the wrong-base comment never suppresses it β€” and the two goroutines race anyway. Result: a second comment saying "The agent did not post its question here … Question: wrong_base: target branch X does not exist", next to the hub's own wrong-base park note. TestNeedsDecisionParkWireShape covers the gate-park release path but not this one.

Suggested fix in contribute_ws.go:

			if verdict == completionVerdictNeedsDecision && completedTask != nil && wrongBaseTarget == "" {

(wrongBaseTarget is non-empty only on the wrong-base path, so this leaves every other needs-decision verdict unchanged.)

2. Minor: the one-minute slack is sent to GitHub (opts.Since = since-1m, contribute_needs_decision_park.go:136-139) but the loop then drops anything Before(since) (:144), so the slack never admits a comment; and when taskAssignedAt is zero (resume-adopted task, contribute_ws.go:154-159) the listing is the unbounded first 100 comments, where any β‰₯40-char comment from any time silently suppresses the post. Both degrade to "nothing posted", so not blocking.

Review findings from ci-maintainer (folded here because this head already carries its hive review):

Intent alignment (read at head 317e891): the diff does what the body says β€” handleTaskComplete fires postNeedsDecisionPark only after the gate-park reclassification (contribute_ws.go:4529-4545), the poster skips non-GitHub / non-issue tasks, lists comments since assignment and posts once behind the <!-- hive:contribute-park --> marker (contribute_needs_decision_park.go:107-152). Prompt/docs/write-site inventory changes are the coupled pieces of the same fix; no scope creep found. One advisory:

Low β€” parkRecommendation newline split is dead in production and over-captures prose. needsDecisionParkComment sanitizes first (contribute_needs_decision_park.go:95), and sanitizeReportedReason collapses all whitespace to single spaces (contribute_reported_reason.go:30), so the \n branch in parkRecommendation (line 73) never fires there β€” only ; splits. A reason like pick a queue owner\nI recommend keeping the scanner (the exact case TestNeedsDecisionParkComment covers, but via a direct parkRecommendation call) becomes one clause, and the whole reason is re-emitted under Recommendation:. Likewise a clause such as ; I have no recommendation is posted as a Recommendation, and the un-park menu's unparkMentionsRecommendation (issue_unpark_menu.go:311-313) will then offer /hive approve for it. Not blocking, since the prompt now asks for the ; recommended: form; suggested tightening so only a labelled clause counts:

func parkRecommendation(reason string) string {
	for _, clause := range strings.FieldsFunc(reason, func(r rune) bool { return r == ';' || r == '\n' }) {
		clause = strings.TrimSpace(clause)
		if recommendationLabelRE.MatchString(clause) {
			return strings.TrimSpace(recommendationLabelRE.ReplaceAllString(clause, ""))
		}
	}
	return ""
}

(and drop the I recommend keeping… assertion in the test, or sanitize before calling it so the test exercises the production path).

Evidence bundle `#16837687b324` sha256:a9691c8bd7b215e2364aecadbe7efac60c100551a8d5fbc955c60b80b9e30d60 β€” download from the dashboard or `hivectl review evidence #16837`.

Review findings from adjudicator-4 (folded here because this head already carries its hive review):

Style re-review at 687b324 (delta from 317e891 is a v6 merge; the only PR-owned line it touched is the upstream log string at src/pkg/dashboard/contribute_ws.go:4505, resolved to match upstream β€” fine).

Prior findings from the earlier head are all still present: the PR's own files are byte-identical apart from that merge line, so the correctness notes on contribute_ws.go:4542 and contribute_needs_decision_park.go:144, the docs note on contributor-relay.md:1627 ("The relay posts no extra comment…"), and the intent note on parkRecommendation at contribute_needs_decision_park.go:73 all carry over unchanged.

One new maintainability point, low severity:

TestNeedsDecisionParkWireShape skips the half the PR body claims it covers β€” src/pkg/dashboard/contribute_needs_decision_park_test.go:246-249. The "a released gate park posts nothing" leg depends on the harness re-offering #77 after the first (needs-decision, full-cooldown) completion, and falls back to t.Skipf when it does not. A skipped leg reports as PASS under plain go test, so the test can't be relied on for the claim in the Tests table. The sibling TestGateParkVerdict_WireShape (contribute_gate_park_test.go:198-217) gets the same assertion deterministically by making the gate-park verdict the first completion of a fresh session and waiting on fx.released. Suggested split:

@@ func TestNeedsDecisionParkWireShape(t *testing.T) {
-	// A gate park is released, not posted.
-	key := noPRKey("acme/widgets", 77)
-	hub.completedMu.Lock()
-	hub.completedTasks[key] = time.Now().Add(-(hub.completedTaskCooldown[key] + time.Minute))
-	delete(hub.noWorkVerdicts, key)
-	hub.completedMu.Unlock()
-	conn.WriteJSON(WSMessage{Type: "ready", Seq: 3})
-	again := readMsg(t, conn)
-	if again.Type != "task_assign" || again.Number != 77 {
-		t.Skipf("issue not re-offered in this harness (%+v); the first half is the regression", again)
-	}
-	conn.WriteJSON(WSMessage{
-		Type: "task_complete", TaskID: again.TaskID, Result: "completed",
-		CompletionSignal: completionSignalVerdict,
-		Verdict:          "no_work_needed", VerdictNeedsDecision: true,
-		VerdictReason: gateParkReason,
-	})
-	select {
-	case c := <-got:
-		t.Fatalf("a released gate park must not get a park comment: %+v", c)
-	case <-time.After(300 * time.Millisecond):
-	}
 }
+
+// A released gate park is not posted: same wire shape as
+// TestGateParkVerdict_WireShape, with the park poster armed.
+func TestNeedsDecisionParkWireShape_GateParkPostsNothing(t *testing.T) {
+	s, ts := setupWSTest(t)
+	defer ts.Close()
+	// …register/dial/auth and seed #77 exactly as above…
+	hub := s.contributeHub
+	got := make(chan call, 1)
+	hub.needsDecisionParkPoster = func(_ context.Context, repo string, number int, since time.Time, body string) error {
+		got <- call{repo, number, since, body}
+		return nil
+	}
+	fx := newToolingReleaseFixture()
+	hub.gateParkReleaser = fx.release
+	conn.WriteJSON(WSMessage{Type: "ready", Seq: 2})
+	assign := readMsg(t, conn)
+	conn.WriteJSON(WSMessage{
+		Type: "task_complete", TaskID: assign.TaskID, Result: "completed",
+		CompletionSignal: completionSignalVerdict,
+		Verdict:          "no_work_needed", VerdictNeedsDecision: true,
+		VerdictReason: gateParkReason,
+	})
+	select {
+	case <-fx.released:
+	case <-time.After(5 * time.Second):
+		t.Fatal("the hub must release the gate park through the seam")
+	}
+	select {
+	case c := <-got:
+		t.Fatalf("a released gate park must not get a park comment: %+v", c)
+	case <-time.After(300 * time.Millisecond):
+	}
+}

(call would move to package scope so both tests share it.)

Otherwise the new file follows contribute_gate_park.go closely β€” same guard set (:111 vs :125), timeout const, seam-with-nil-means-GHClient, sanitizeReportedReason, and the // writeback:solicited annotation at :151 β€” so no further style concerns.

Review findings from reviewer (folded here because this head already carries its hive review):

Read the new file against the v6 tree and the handleTaskComplete wiring. The change looks correct to me: taskAssignedAt is copied at contribute_ws.go:4341 before the session zeroes it at :4363, so the detached goroutine gets the real assignment time; the park call sits after reclassifyGateParkVerdict (:4523-4531) and keys on the post-reclassification verdict, so a released gate park is never commented on; and sanitizeReportedReason (contribute_reported_reason.go:26-32) strips HTML-comment delimiters to a fixed point, so a reason cannot forge the <!-- hive:contribute-park task=… --> marker the idempotency check keys on. Two low notes:

  1. The "minute of slack" is undone by the filter. needsDecisionParkPost asks GitHub for comments since assignedAt-1m (contribute_needs_decision_park.go:135-138) but then only counts comments with !CreatedAt.Before(since) (:143), i.e. created at or after the unadjusted assignedAt. The slack therefore widens the query without ever affecting the decision. If the hub clock runs a few seconds ahead of GitHub's created_at, an agent explanation posted right after assignment is excluded and the hub posts a second question. Either apply the slack in the filter too, or drop the comment that says there is slack.

  2. parkRecommendation matches negated clauses. Any ;-clause containing recommend is rendered as **Recommendation:** … (:71-78). A reason such as decision: A or B; I have no recommendation produces **Recommendation:** I have no recommendation β€” which, per your own note on the un-park menu, is the case where /hive approve would offer an empty choice. Requiring the label form your prompt now asks for (recommended: / recommendation: prefix, i.e. recommendationLabelRE must actually match) would avoid it.

Neither blocks landing; both are a few lines.

@hivecommons-hive

Copy link
Copy Markdown
Contributor

@Danathar this PR is back in your court (stage/072-returned-to-contributor). The hive will not merge it until the head changes.

Review asks:

  • the review swarm requested changes

Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path.

Queued in your relay fix queue (9 waiting): fixes are handed out before new work as soon as you have capacity.

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Delta re-review at 687b324 (correctness). The new head is only Merge branch 'v6'; the sole change to a PR-owned file is an upstream log string in contribute_ws.go:4505. contribute_needs_decision_park.go is byte-identical to 317e891, so the two previous correctness findings are still present:

1. Wrong-base fallback posts two hub park comments β€” contribute_ws.go:4488-4496 turns an unverifiable wrong_base into needs_decision and goes parkWrongBaseIssue, which posts a <!-- hive:wrong-base-park --> comment (contribute_wrong_base.go:14,117). The same completion then reaches contribute_ws.go:4542-4545 and goes postNeedsDecisionPark. Even when the wrong-base comment lands first, parkCommentExplains (contribute_needs_decision_park.go:54-56) returns false for any <!-- hive: notice, so the question is posted a second time. Suggested fix in contribute_ws.go:

 			wrongBaseTarget := ""
+			wrongBaseParked := false
 			if verdict == completionVerdictWrongBase {
 				resolved, target, reason, park := h.resolveWrongBaseVerdict(
 					completedTask, verdict, msg.VerdictTargetBranch)
 				verdict, wrongBaseTarget, verdictReason = resolved, target, reason
+				wrongBaseParked = park
 				if park && completedTask != nil {
@@
-			if verdict == completionVerdictNeedsDecision && completedTask != nil {
+			if verdict == completionVerdictNeedsDecision && completedTask != nil && !wrongBaseParked {

2. Slack is a no-op; zero assignedAt makes the scan unbounded β€” contribute_needs_decision_park.go:160-163 asks GitHub for comments since since-1m, but line 167 then drops anything Before(since), so the minute of slack never admits a comment. When since is zero (resume-adopted tasks, contribute_ws.go:157-159) no Since is sent and line 167 accepts every one of the first 100 comments, so any β‰₯40-char comment in the issue's history (the original report, a triage note) suppresses the question β€” exactly the unexplained park this PR exists to fix. Suggested fix:

 	opts := &gh.IssueListCommentsOptions{ListOptions: gh.ListOptions{PerPage: 100}}
-	if !since.IsZero() {
-		at := since.Add(-time.Minute)
-		opts.Since = &at
+	if since.IsZero() {
+		return fmt.Errorf("task has no assignment time; not scanning the whole issue for an explanation")
 	}
+	at := since.Add(-time.Minute)
+	opts.Since = &at
@@
 	for _, c := range comments {
-		if since.IsZero() || !c.GetCreatedAt().Time.Before(since) {
+		if !c.GetCreatedAt().Time.Before(at) {
 			if parkCommentExplains(c.GetBody()) {
 				return nil
 			}
 		}
 	}

(Either skip-with-log as above, or post unconditionally when since is zero β€” but then dedupe needs the marker check to still run.)

Previous docs-currency note (contributor-relay.md:1624) is fixed at this head β€” the doc now describes the hub comment. The intent-alignment note on parkRecommendation (contribute_needs_decision_park.go:69-78) is unchanged and still applies.

β€” hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88

@hivecommons-hive
hivecommons-hive Bot merged commit 03bf5d7 into hivecommons:v6 Nov 6, 2026
69 of 70 checks passed
@hivecommons-hive

Copy link
Copy Markdown
Contributor

merged by Hive trusted-author auto-merge: author @Danathar holds role owner and GitHub permission maintain on repo; CI green at 687b324.

Danathar added a commit to Danathar/hive that referenced this pull request Nov 6, 2026
Review on hivecommons#16837:
- quality: only the reporting contributor's own comment (or the hub's
  marker) counts as the park's explanation; a maintainer's or bystander's
  remark in the task window no longer suppresses the question. The wire
  test's gate-park half now fails instead of skipping, and
  needsDecisionParkDue is tested directly.
- adjudicator-3: a wrong_base verdict that falls back to needs-decision
  already gets parkWrongBaseIssue's note; it no longer also gets the hub's
  question (an explicit flag, since the target can be empty). The
  one-minute clock-skew slack now applies to the filter as well as the
  listing.
- reviewer-2: contributor-relay.md's "no extra comment" sentence names the
  hub fallback.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1
Signed-off-by: Danathar <Danathar@users.noreply.github.com>
@Danathar

Danathar commented Nov 6, 2026

Copy link
Copy Markdown
Contributor Author

All four review findings are fixed at ad5afd1 (fix commit d34edc7, then merges of v6):

  • quality, author: only the reporting contributor's own comment, or the hub's own marker, counts as the park's explanation (parkCommentExplains(body, author, reporter)). A maintainer's or bystander's remark in the task window no longer suppresses the question. New rows cover a third-party comment (still posts) and the reporter's explanation (skips).
  • quality, Skipf: the wire test's gate-park half now fails instead of skipping. TestNeedsDecisionParkDue also asserts directly that a demoted gate park is not due.
  • adjudicator-3 🌱 Sync workflows from kubestellar/infraΒ #1, wrong-base double comment: a wrong_base verdict that falls back to needs-decision no longer gets the hub's question next to parkWrongBaseIssue's note. I used an explicit wrongBaseParked flag instead of wrongBaseTarget == "", because the no-valid-target fallback parks with an empty target. TestNeedsDecisionParkNotForWrongBaseFallback drives that path through the real handler.
  • adjudicator-3 docs: add EXECUTOR MODE pattern and multi-agent topologyΒ #2, slack: the one-minute clock-skew slack now applies to the filter as well as the listing. With the author check, an unbounded listing (zero assignment time) can only be suppressed by the reporter's own comments.
  • reviewer-2, docs: applied your suggested sentence in contributor-relay.md.

Mutants on the new logic: 7/7 caught (17/17 overall). go test for dashboard, github and writeback passes, and so does the docs citation check.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dco-signoff: yes Indicates the PR's author has signed the DCO. size/L Denotes a PR that changes 100-499 lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant