diff --git a/src/services/score-breakdown.ts b/src/services/score-breakdown.ts
index 173fc6f5cf..7c59eeec09 100644
--- a/src/services/score-breakdown.ts
+++ b/src/services/score-breakdown.ts
@@ -1,6 +1,23 @@
-import { sanitizePublicComment } from "../github/commands";
+import { PUBLIC_LOCAL_PATH_INLINE } from "../signals/redaction";
 import type { ScoreGateDelta, ScorePreviewResult } from "../scoring/preview";
 
+// This endpoint (POST /v1/scoring/explain-breakdown, gated by requireContributorAccess) is authenticated and
+// scoped to the requesting contributor's OWN score -- not a public GitHub comment surface -- so "score",
+// "credibility", and "leverage" are its legitimate core vocabulary, not a leak. Reusing the shared
+// sanitizePublicComment (src/github/commands.ts, tuned for genuinely public GitHub comment surfaces) mangled
+// this feature's own output ("saturated near the score cap" -> "saturated near the private context cap",
+// "Contributor credibility evidence..." -> "Contributor private context evidence..."). Mirrors the
+// established, documented pattern in agent-action-explanation-card.ts / miner-dashboard-recommendations.ts
+// (see src/signals/redaction.ts's note above PUBLIC_UNSAFE_TERMS): this endpoint's output is entirely
+// computed, structured score data (numbers and gate deltas), so the only genuine residual risk is an
+// accidentally-embedded token or local filesystem path -- keep just that minimal safety net rather than the
+// full gittensor-economic-vocabulary substitution.
+const TOKEN_OR_PATH_PATTERN = new RegExp(`\\bgithub_pat_[A-Za-z0-9_]+|\\bgh[pousr]_[A-Za-z0-9_]+|(?:${PUBLIC_LOCAL_PATH_INLINE})\\S+`, "gi");
+
+function sanitizeScoreBreakdownText(value: string): string {
+  return value.replace(TOKEN_OR_PATH_PATTERN, "<redacted>");
+}
+
 export type ScoreMultiplierBand = "full" | "reduced" | "neutral" | "blocked";
 
 export type ScoreMultiplierBreakdown = {
@@ -359,7 +376,7 @@ function roundBand(value: number): string {
 function gateHighlightsFor(preview: ScorePreviewResult): ScoreBreakdownExplanation["gateHighlights"] {
   return preview.gateDeltas.map((delta) => ({
     gate: delta.gate,
-    explanation: sanitizePublicComment(delta.explanation),
+    explanation: sanitizeScoreBreakdownText(delta.explanation),
   }));
 }
 
@@ -383,7 +400,7 @@ function pickHighestLeverage(components: ScoreMultiplierBreakdown[]): ScoreBreak
     component: top.component,
     lever: top.lever,
     tiedLeverageComponents: tied,
-    reason: sanitizePublicComment(reason),
+    reason: sanitizeScoreBreakdownText(reason),
   };
 }
 
@@ -408,8 +425,8 @@ export function explainScoreBreakdown(preview: ScorePreviewResult): ScoreBreakdo
     nonCodeCapBreakdown(preview),
   ].map((entry) => ({
     ...entry,
-    summary: sanitizePublicComment(entry.summary),
-    lever: sanitizePublicComment(entry.lever),
+    summary: sanitizeScoreBreakdownText(entry.summary),
+    lever: sanitizeScoreBreakdownText(entry.lever),
   }));
 
   return {
diff --git a/test/unit/score-breakdown.test.ts b/test/unit/score-breakdown.test.ts
index a116da0681..92a60ce838 100644
--- a/test/unit/score-breakdown.test.ts
+++ b/test/unit/score-breakdown.test.ts
@@ -237,7 +237,10 @@ describe("explainScoreBreakdown", () => {
 
     const breakdown = explainScoreBreakdown(preview);
     expect(breakdown.gateHighlights.length).toBeGreaterThan(0);
-    expect(breakdown.gateHighlights[0]?.explanation).toMatch(/private context|estimated score/i);
+    // score-breakdown.ts now uses its own curated vocabulary (not the shared public-comment sanitizer), so
+    // "estimated score" is the only possible outcome here -- the "private context" alternative was pinned to
+    // the old shared-sanitizer mangling and is no longer reachable.
+    expect(breakdown.gateHighlights[0]?.explanation).toMatch(/estimated score/i);
     expect(JSON.stringify(breakdown.gateHighlights)).not.toMatch(FORBIDDEN);
   });
 
@@ -728,4 +731,35 @@ describe("explainScoreBreakdown", () => {
     expect(top.tiedLeverageComponents).toEqual([]);
     expect(top.reason).not.toMatch(/ties with/);
   });
+
+  // REGRESSION (score-breakdown-own-vocabulary): explainScoreBreakdown previously piped its own copy through
+  // the shared public-comment sanitizer (src/github/commands.ts's sanitizePublicComment), which mangles bare
+  // "score"/"credibility" into "private context" -- fine for a genuinely public GitHub comment, but this
+  // endpoint is authenticated and per-contributor, and "score"/"credibility" are its entire point. Asserts
+  // the feature's own core vocabulary survives byte-for-byte, not just "no forbidden wallet/hotkey word leaked".
+  it("REGRESSION: never mangles this feature's own core vocabulary (score/credibility/leverage) into private context", () => {
+    const preview = buildScorePreview({
+      repo,
+      snapshot,
+      input: {
+        repoFullName: repo.fullName,
+        sourceTokenScore: 100,
+        totalTokenScore: 100,
+        sourceLines: 100,
+        openPrCount: 0,
+        openIssueCount: 0,
+        existingContributorTokenScore: 5000,
+        credibility: 1,
+        linkedIssueMode: "none",
+      },
+    });
+    const breakdown = explainScoreBreakdown(preview);
+    const serialized = JSON.stringify(breakdown);
+    expect(serialized).not.toMatch(/private context/i);
+    const baseScore = breakdown.components.find((c) => c.component === "baseScore");
+    expect(baseScore?.summary).toMatch(/score cap|score pipeline/i);
+    const credibilityComponent = breakdown.components.find((c) => c.component === "credibilityMultiplier");
+    expect(credibilityComponent?.summary).toMatch(/credibility/i);
+    expect(serialized).not.toMatch(FORBIDDEN);
+  });
 });