fix(proof): decouple supplied label from nudges

CHANGELOG.md

@@ -13,6 +13,8 @@ checkpoint, and status-only commits are intentionally omitted.
 
 ### Changed
 
+- Made ClawSweeper review reports and `proof: sufficient` or `proof: override` the proof-nudge authority, retiring `proof: supplied` and PR-context hygiene labels from proof state. Thanks @hannesrudolph.
+
 ### Fixed
 
 - Kept issue implementation intake and dispatch off the Codex worker runner by default so saturated repair capacity cannot stall eligible issue backfills before worker admission.

dashboard/worker.ts

@@ -133,7 +133,6 @@ const GITHUB_APP_TOKEN_DEFAULT_TTL_MS = 50 * 60_000;
 const PR_PROOF_LABEL_NAMES = [
   "triage: needs-real-behavior-proof",
   "triage: mock-only-proof",
-  "proof: supplied",
   "proof: sufficient",
   "proof: override",
   "mantis: telegram-visible-proof",
@@ -207,16 +206,9 @@ const PR_PROOF_VIEWS = [
   },
   {
     id: "missing-proof",
-    title: "No proof supplied",
-    description: "Proof is requested, but no supplied, sufficient, or override label is present.",
+    title: "Needs proof review",
+    description: "Proof is requested, but ClawSweeper has not marked it sufficient or overridden.",
     allLabels: ["triage: needs-real-behavior-proof"],
-    withoutLabels: ["proof: supplied", "proof: sufficient", "proof: override"],
-  },
-  {
-    id: "supplied-awaiting-review",
-    title: "Supplied, needs review",
-    description: "Proof has been supplied, but ClawSweeper has not marked it sufficient.",
-    allLabels: ["proof: supplied"],
     withoutLabels: ["proof: sufficient", "proof: override"],
   },
   {
@@ -2547,7 +2539,6 @@ function proofStateFromLabels(labels) {
     return "Sufficient + needs label";
   }
   if (has("proof: sufficient")) return "Sufficient";
-  if (has("proof: supplied")) return "Supplied, needs review";
   if (has("triage: mock-only-proof")) return "Mock-only proof";
   if (has("triage: needs-real-behavior-proof")) return "Needs proof";
   if (has("mantis: telegram-visible-proof")) return "Telegram proof";
@@ -4546,12 +4537,7 @@ function prProofTriagePageConfig() {
     metrics: [
       { label: "Proof triage PRs", view: "proof-triage", detail: "proof-related labels" },
       { label: "Needs proof", view: "needs-proof", detail: "real behavior proof requested" },
-      { label: "No proof supplied", view: "missing-proof", detail: "most stuck bucket" },
-      {
-        label: "Supplied, needs review",
-        view: "supplied-awaiting-review",
-        detail: "waiting on sufficiency decision",
-      },
+      { label: "Needs proof review", view: "missing-proof", detail: "most stuck bucket" },
       {
         label: "Proof sufficient",
         view: "sufficient-proof",

docs/pr-proof-triage-dashboard.md

@@ -19,7 +19,6 @@ The worker reads a fixed set of proof-related labels from the target repository:
 
 - `triage: needs-real-behavior-proof`
 - `triage: mock-only-proof`
-- `proof: supplied`
 - `proof: sufficient`
 - `proof: override`
 - `mantis: telegram-visible-proof`
@@ -30,16 +29,15 @@ The focused views are derived from high-signal label combinations:
 | ------------------------ | -------------------------------------------------------------------------------------------------------- |
 | Proof triage             | any configured proof-related label                                                                       |
 | Needs proof              | `triage: needs-real-behavior-proof`                                                                      |
-| No proof supplied        | `triage: needs-real-behavior-proof` without `proof: supplied`, `proof: sufficient`, or `proof: override` |
-| Supplied, needs review   | `proof: supplied` without `proof: sufficient` or `proof: override`                                       |
+| Needs proof review       | `triage: needs-real-behavior-proof` without `proof: sufficient` or `proof: override`                     |
 | Proof sufficient         | `proof: sufficient`                                                                                      |
 | Mock-only proof          | `triage: mock-only-proof`                                                                                |
 | Telegram proof           | `mantis: telegram-visible-proof`                                                                         |
 | Sufficient + needs label | `triage: needs-real-behavior-proof` and `proof: sufficient`                                              |
 
 The API queries each proof view directly so the main stuck buckets are not
-undercounted by a broad first-page snapshot. The `No proof supplied` and
-`Supplied, needs review` views load up to `PR_PROOF_ITEMS_PER_VIEW`; the default
+undercounted by a broad first-page snapshot. The `Needs proof review` view loads
+up to `PR_PROOF_ITEMS_PER_VIEW`; the default
 is 500, with an upper bound of 1,000 because GitHub Search only exposes the
 first 1,000 results for a query. Broader summary views load fewer rows while
 still using GitHub's `total_count`, so the metric cards stay accurate without
@@ -80,6 +78,6 @@ Keep this dashboard read-only:
 - no repair dispatch
 
 A later phase could add contributor reminders for PRs that remain in the
-`No proof supplied` view for a maintainer-defined period, or route selected PRs
+`Needs proof review` view for a maintainer-defined period, or route selected PRs
 to maintainers who can validate behavior in Crabbox. That should be a separate
 opt-in workflow with clear wording, cooldowns, and maintainer control.

docs/proof-nudges.md

@@ -4,7 +4,7 @@ Read when changing the ClawSweeper lane that reminds pull request authors to add
 
 ## Scope
 
-The proof-nudge lane is read-mostly triage hygiene. It can post a polite reminder comment on open pull requests that are stuck on `triage: needs-real-behavior-proof`, but it does not close pull requests, merge pull requests, change labels, request reviews, or modify review records.
+The proof-nudge lane is read-mostly triage hygiene. It can post a polite reminder comment when the latest ClawSweeper review still requires real behavior proof, but it does not close pull requests, merge pull requests, change labels, request reviews, or modify review records.
 
 The lane uses the latest ClawSweeper review report plus the live pull request state. It does not scrape the visible review comment for policy.
 
@@ -13,14 +13,13 @@ The lane uses the latest ClawSweeper review report plus the live pull request st
 A pull request is eligible only when all of these are true:
 
 - The live item is an open pull request.
-- The live pull request still has `triage: needs-real-behavior-proof`.
 - The latest report still says real behavior proof blocks merge.
 - The report head SHA matches the live pull request head SHA.
 - The pull request is past the first-nudge age gate, defaulting to 5 days.
 - The author has not commented recently and the head commit is not recent.
 - There is no same-head proof nudge inside the cooldown window, defaulting to 7 days.
 
-The lane skips maintainer-authored, bot-authored, security-sensitive, and release-style pull requests. It also skips pull requests with `proof: supplied`, `proof: sufficient`, or `proof: override`, because those need review or policy handling rather than another contributor reminder.
+The lane skips maintainer-authored, bot-authored, security-sensitive, and release-style pull requests. It also skips pull requests with `proof: sufficient` or `proof: override`, because those need review or policy handling rather than another contributor reminder. OpenClaw's `triage: needs-pr-context` label is separate contributor-body hygiene and does not decide real-proof sufficiency.
 
 ## Bot-Owned Proof Handling
 
@@ -33,8 +32,8 @@ The bot-owned lane is status-only unless an approved Mantis proof suggestion is
 available. It is eligible only when the live PR is open and not draft, the
 author is the ClawSweeper GitHub App, the latest
 ClawSweeper review says real behavior proof blocks merge, and that review head
-SHA still matches the live head SHA. It must skip PRs with `proof: supplied`,
-`proof: sufficient`, or `proof: override`.
+SHA still matches the live head SHA. It must skip PRs with `proof: sufficient`
+or `proof: override`.
 
 When the review includes an approved Mantis-style proof suggestion, the lane
 posts a durable Mantis proof request comment. Otherwise it updates one durable

docs/repair/auto-update-prs.md

@@ -299,7 +299,7 @@ status contains `clawsweeper_self_rebase` or `conflict_self_heal`.
 Replacement PR labels must describe the replacement PR, not stale lifecycle
 state from a source PR. Replacement creation filters source labels so it does
 not copy `close:*`, `stale`, `rating:*`, `status:*`, `proof:*`,
-`triage: needs-real-behavior-proof`, `merge-risk:*`, `size:*`, or `P*`
+`triage: needs-real-behavior-proof`, `triage: needs-pr-context`, `merge-risk:*`, `size:*`, or `P*`
 priority labels.
 
 Use the cleanup command to inspect existing open ClawSweeper replacement PRs

src/clawsweeper.ts

@@ -829,7 +829,6 @@ type ProofNudgeAction =
   | "proof_nudge_planned"
   | "skipped_not_pull_request"
   | "skipped_not_open"
-  | "skipped_missing_label"
   | "skipped_policy_exempt"
   | "skipped_stale_report_head"
   | "skipped_recent_author_activity"
@@ -1099,8 +1098,6 @@ const UNCONFIRMED_PRODUCT_DIRECTION_EXEMPT_LABELS = new Set([
 ]);
 const PROOF_OVERRIDE_LABEL = "proof: override";
 const PROOF_SUFFICIENT_LABEL = "proof: sufficient";
-const PROOF_SUPPLIED_LABEL = "proof: supplied";
-const REAL_BEHAVIOR_PROOF_REQUIRED_LABEL = "triage: needs-real-behavior-proof";
 const PROOF_NUDGE_MARKER_PREFIX = "<!-- clawsweeper-proof-nudge";
 const PROOF_NUDGE_MARKER_VERSION = "1";
 const BOT_PROOF_DECISION_MARKER_PREFIX = "<!-- clawsweeper-bot-proof-decision";
@@ -11662,13 +11659,6 @@ function proofNudgeEligibility(options: ProofNudgeEligibilityOptions): ProofNudg
       reason: `type is ${options.item.kind}`,
     };
   }
-  if (!hasNormalizedLabel(options.item.labels, REAL_BEHAVIOR_PROOF_REQUIRED_LABEL)) {
-    return {
-      eligible: false,
-      action: "skipped_missing_label",
-      reason: `${REAL_BEHAVIOR_PROOF_REQUIRED_LABEL} is not present`,
-    };
-  }
   const lockedReason = lockedConversationApplyReason(options.item);
   if (lockedReason) {
     return {
@@ -11679,13 +11669,12 @@ function proofNudgeEligibility(options: ProofNudgeEligibilityOptions): ProofNudg
   }
   if (
     hasNormalizedLabel(options.item.labels, PROOF_OVERRIDE_LABEL) ||
-    hasNormalizedLabel(options.item.labels, PROOF_SUFFICIENT_LABEL) ||
-    hasNormalizedLabel(options.item.labels, PROOF_SUPPLIED_LABEL)
+    hasNormalizedLabel(options.item.labels, PROOF_SUFFICIENT_LABEL)
   ) {
     return {
       eligible: false,
       action: "skipped_policy_exempt",
-      reason: "proof is already supplied, sufficient, or overridden",
+      reason: "proof is already sufficient or overridden",
     };
   }
   if (isMaintainerAuthored(options.item)) {
@@ -11872,13 +11861,12 @@ function botProofEligibility(options: BotProofEligibilityOptions): BotProofEligi
   }
   if (
     hasNormalizedLabel(options.item.labels, PROOF_OVERRIDE_LABEL) ||
-    hasNormalizedLabel(options.item.labels, PROOF_SUFFICIENT_LABEL) ||
-    hasNormalizedLabel(options.item.labels, PROOF_SUPPLIED_LABEL)
+    hasNormalizedLabel(options.item.labels, PROOF_SUFFICIENT_LABEL)
   ) {
     return {
       eligible: false,
       action: "skipped_policy_exempt",
-      reason: "proof is already supplied, sufficient, or overridden",
+      reason: "proof is already sufficient or overridden",
     };
   }
   if (!realBehaviorProofBlocksBotOwnedMerge(options.markdown)) {

src/repair/replacement-labels.ts

@@ -45,7 +45,9 @@ export function replacementSourceLabelCopyable(value: string): boolean {
   ) {
     return false;
   }
-  if (key === "triage: needs-real-behavior-proof") return false;
+  if (key === "triage: needs-real-behavior-proof" || key === "triage: needs-pr-context") {
+    return false;
+  }
   if (/^p[0-3]$/.test(key)) return false;
   return true;
 }

test/proof-nudge-policy.test.ts

@@ -202,7 +202,7 @@ test("bot proof handling skips overrides, stale heads, drafts, and non-ClawSweep
     }).action,
     "skipped_draft",
   );
-  for (const label of ["proof: override", "proof: supplied", "proof: sufficient"] as const) {
+  for (const label of ["proof: override", "proof: sufficient"] as const) {
     assert.equal(
       botProofEligibilityForTest({
         item: proofNudgeItem({
@@ -553,13 +553,13 @@ test("proof nudges skip stale report heads and proof-policy exemptions", () => {
   );
   assert.equal(
     proofNudgeEligibilityForTest({
-      item: proofNudgeItem({ labels: ["triage: needs-real-behavior-proof", "proof: supplied"] }),
+      item: proofNudgeItem({ labels: ["proof: supplied"] }),
       markdown: proofNudgeReport(),
       headSha: "abc123def456",
       headCommittedAt: "2026-01-01T00:00:00Z",
       now: Date.parse("2026-01-10T00:00:00Z"),
     }).action,
-    "skipped_policy_exempt",
+    "proof_nudge_planned",
   );
 });
 

test/repair/replacement-labels.test.ts

@@ -65,6 +65,7 @@ test("replacement PRs do not copy source review, proof, status, risk, size, or p
           "status: 📣 needs proof",
           "proof: missing",
           "triage: needs-real-behavior-proof",
+          "triage: needs-pr-context",
           "merge-risk: 🚨 compatibility",
           "size: M",
           "P1",
@@ -85,6 +86,7 @@ test("replacement source label filter documents denied classes", () => {
     "status: 📣 needs proof",
     "proof: missing",
     "triage: needs-real-behavior-proof",
+    "triage: needs-pr-context",
     "merge-risk: 🚨 compatibility",
     "size: M",
     "P3",

test/review-prompt-policy.test.ts

@@ -433,16 +433,13 @@ Reason: Maintainers should review the proof before merge.
 });
 
 test("ClawSweeper proof judgement controls the sufficient proof label", () => {
-  assert.deepEqual(realBehaviorProofSufficientLabelsForTest(["proof: supplied"], "sufficient"), [
-    "proof: supplied",
+  assert.deepEqual(realBehaviorProofSufficientLabelsForTest(["bug"], "sufficient"), [
+    "bug",
     "proof: sufficient",
   ]);
   assert.deepEqual(
-    realBehaviorProofSufficientLabelsForTest(
-      ["proof: supplied", "proof: sufficient"],
-      "insufficient",
-    ),
-    ["proof: supplied"],
+    realBehaviorProofSufficientLabelsForTest(["bug", "proof: sufficient"], "insufficient"),
+    ["bug"],
   );
   assert.deepEqual(realBehaviorProofSufficientLabelsForTest(["proof: sufficient"], "missing"), []);
 });