feat(review): capture bounded raw context in the linked_issue_scope_mismatch fired signal

src/queue/processors.ts

@@ -492,6 +492,7 @@ import { buildSlopAssessment, type SlopBand } from "../signals/slop";
 import { copycatWouldActOnPersistedScore } from "../signals/copycat";
 import { buildStructuralImprovementAssessment } from "../signals/improvement";
 import { runLoopOverLinkedIssueSatisfaction } from "../services/linked-issue-satisfaction-run";
+import { MAX_BODY_CHARS, MAX_DIFF_CHARS, MAX_ISSUE_TEXT_CHARS } from "../services/linked-issue-satisfaction";
 import { decidePublicSurface } from "../signals/settings-preview";
 import {
   buildFocusManifestGuidance,
@@ -7553,7 +7554,18 @@ export async function runLinkedIssueSatisfactionForAdvisory(
           targetKey: `${args.repoFullName}#${args.pr.number}`,
           outcome: result.result.status,
           occurredAt: nowIso(),
-          metadata: { confidence: result.result.confidence },
+          // #8129: also capture the bounded raw inputs the assessment was based on — a future backtest
+          // classify() re-runs new prompt/logic against these and compares to the recorded label; confidence
+          // alone can only validate threshold changes. Bounds mirror LinkedIssueSatisfactionInput's own
+          // (reused constants + the exact trim/slice shapes buildLinkedIssueSatisfactionPrompt applies), so
+          // what's stored is byte-what the assessment actually saw.
+          metadata: {
+            confidence: result.result.confidence,
+            issueText: issueText.trim().slice(0, MAX_ISSUE_TEXT_CHARS),
+            prTitle: args.pr.title,
+            prBody: (args.pr.body ?? "").trim().slice(0, MAX_BODY_CHARS),
+            diff: diff.slice(0, MAX_DIFF_CHARS),
+          },
         })
         .catch(() => undefined);
     }

src/services/linked-issue-satisfaction.ts

@@ -29,9 +29,11 @@ export type LinkedIssueSatisfactionStatus = (typeof LINKED_ISSUE_SATISFACTION_ST
 export const LINKED_ISSUE_SATISFACTION_CONFIDENCE_FLOOR = 0.5;
 
 const MAX_RATIONALE_LENGTH = 400;
-const MAX_ISSUE_TEXT_CHARS = 6000;
-const MAX_DIFF_CHARS = 60000;
-const MAX_BODY_CHARS = 2000;
+// Exported (#8129) so the calibration fired-event capture reuses the assessment's OWN bounds for the raw
+// context it stores, instead of drifting behind a second set of hand-maintained limits.
+export const MAX_ISSUE_TEXT_CHARS = 6000;
+export const MAX_DIFF_CHARS = 60000;
+export const MAX_BODY_CHARS = 2000;
 
 export type LinkedIssueSatisfactionInput = {
   /** The already-fetched linked-issue title + body text (grounding already resolved this — see

test/unit/linked-issue-satisfaction-run.test.ts

@@ -1,5 +1,6 @@
 import { afterEach, describe, expect, it, vi } from "vitest";
 import { runLoopOverLinkedIssueSatisfaction, type LinkedIssueSatisfactionRunInput } from "../../src/services/linked-issue-satisfaction-run";
+import { MAX_BODY_CHARS, MAX_DIFF_CHARS } from "../../src/services/linked-issue-satisfaction";
 import { BEST_REVIEW_MODELS, RELIABLE_FALLBACK_MODELS } from "../../src/services/ai-review";
 import { buildAiReviewDiff, processJob, runLinkedIssueSatisfactionForAdvisory } from "../../src/queue/processors";
 import { evaluateGateCheck } from "../../src/rules/advisory";
@@ -481,6 +482,38 @@ describe("runLinkedIssueSatisfactionForAdvisory (processor wiring, #1961/#3906)"
       expect(history.overrides).toEqual([]); // firing alone is never an override
     });
 
+    it("captures the bounded raw context (issueText/prTitle/prBody/diff) on the fired signal, truncating an over-limit body (#8129)", async () => {
+      stubIssueFetch();
+      const run = vi.fn(async () => ({ response: satisfactionJson({ status: "unaddressed", confidence: 0.9 }) }));
+      const env = enabledEnv(run);
+      const longBody = "x".repeat(MAX_BODY_CHARS + 500);
+      const longBodyPr = { ...pr, body: longBody };
+      await runLinkedIssueSatisfactionForAdvisory(env, { mode: "live", settings: blockMode, advisory: advisory(), repoFullName: "acme/widgets", pr: longBodyPr, author: "alice", files, confirmedContributor: true, installationId: 1 });
+
+      const history = await createSignalStore(env).queryRuleHistory("linked_issue_scope_mismatch", 0);
+      expect(history.fired).toHaveLength(1);
+      const metadata = history.fired[0]?.metadata as Record<string, string>;
+      // The stored issue text is the same title+body composite the assessment itself consumed.
+      expect(metadata.issueText).toContain("Enrich SN74 Gittensor — add SSE stream");
+      expect(metadata.issueText).toContain("We need a live SSE stream surface for SN74 Gittensor.");
+      expect(metadata.prTitle).toBe("Add SSE stream endpoint");
+      // The over-limit body is truncated to the assessment's OWN bound, never stored raw.
+      expect(metadata.prBody).toBe(longBody.slice(0, MAX_BODY_CHARS));
+      expect(metadata.prBody).toHaveLength(MAX_BODY_CHARS);
+      expect(metadata.diff).toBe(buildAiReviewDiff(files).slice(0, MAX_DIFF_CHARS));
+    });
+
+    it("stores an empty prBody (not the string 'null'/'undefined') for a body-less PR (#8129)", async () => {
+      stubIssueFetch();
+      const run = vi.fn(async () => ({ response: satisfactionJson({ status: "unaddressed", confidence: 0.9 }) }));
+      const env = enabledEnv(run);
+      const bodylessPr = { ...pr, body: null as unknown as string };
+      await runLinkedIssueSatisfactionForAdvisory(env, { mode: "live", settings: blockMode, advisory: advisory(), repoFullName: "acme/widgets", pr: bodylessPr, author: "alice", files, confirmedContributor: true, installationId: 1 });
+
+      const history = await createSignalStore(env).queryRuleHistory("linked_issue_scope_mismatch", 0);
+      expect((history.fired[0]?.metadata as Record<string, string>).prBody).toBe("");
+    });
+
     it("ADVISORY mode records NO fired signal for the same 'unaddressed' verdict (#8101 — no finding, no signal)", async () => {
       stubIssueFetch();
       const run = vi.fn(async () => ({ response: satisfactionJson({ status: "unaddressed", confidence: 0.9 }) }));