fix(filters): match co-op/intern skip tokens in titles with hyphen normalization

Company skip list and profile deal-breakers now ignore hyphens when matching,
apply skip keywords to job titles as well as employers, and short-circuit
scoring when a deal-breaker hits so co-op posts cannot score 80 via the LLM.

Co-authored-by: Cursor <cursoragent@cursor.com>

orchestrator/src/client/pages/orchestrator/useFilteredJobs.test.ts

@@ -454,6 +454,37 @@ describe("useFilteredJobs", () => {
     );
     expect(settingsSkip.current.map((j) => j.id)).toEqual(["good"]);
 
+    const { result: coOpTitle } = renderHook(() =>
+      useFilteredJobs(
+        [
+          {
+            ...baseJob,
+            id: "coop",
+            employer: "Global Relay Communications Inc.",
+            title: "Co-op Software Development Engineer in Test (SDET)",
+          },
+          { ...baseJob, id: "good", employer: "Contoso", title: "SDET" },
+        ],
+        "all",
+        [],
+        [],
+        [],
+        [],
+        "all",
+        "all",
+        { mode: "at_least", min: null, max: null },
+        { key: "score", direction: "desc" },
+        {
+          foundAfterYmd: null,
+          foundBeforeYmd: null,
+          employerInclude: [],
+          employerExclude: [],
+          settingsBlockedEmployerKeywords: ["co-op"],
+        },
+      ),
+    );
+    expect(coOpTitle.current.map((j) => j.id)).toEqual(["good"]);
+
     const { result: includeOnly } = renderHook(() =>
       useFilteredJobs(
         jobs,

orchestrator/src/client/pages/orchestrator/useFilteredJobs.ts

@@ -1,3 +1,4 @@
+import { textMatchesKeyword } from "@shared/keyword-match";
 import { inferCountryKeysFromJobLocation } from "@shared/search-cities";
 import type { JobListItem, JobSource } from "@shared/types";
 import { useMemo } from "react";
@@ -24,13 +25,16 @@ const startOfLocalDayMs = (ymd: string): number =>
 const endOfLocalDayMs = (ymd: string): number =>
   new Date(`${ymd}T23:59:59.999`).getTime();
 
-function employerMatchesAnyKeyword(
-  employer: string,
-  keywordsLower: string[],
+function jobMatchesAnyBlockedKeyword(
+  job: Pick<JobListItem, "employer" | "title">,
+  keywords: string[],
 ): boolean {
-  if (keywordsLower.length === 0) return false;
-  const normalized = employer.toLowerCase();
-  return keywordsLower.some((keyword) => normalized.includes(keyword));
+  if (keywords.length === 0) return false;
+  return keywords.some(
+    (keyword) =>
+      textMatchesKeyword(job.employer, keyword) ||
+      textMatchesKeyword(job.title, keyword),
+  );
 }
 
 const getSponsorCategory = (score: number | null): SponsorFilter => {
@@ -63,12 +67,10 @@ export const useFilteredJobs = (
     const employerIncludeLower = listExtras.employerInclude.map((value) =>
       value.toLowerCase(),
     );
-    const employerExcludeLower = [
-      ...listExtras.settingsBlockedEmployerKeywords.map((value) =>
-        value.toLowerCase(),
-      ),
-      ...listExtras.employerExclude.map((value) => value.toLowerCase()),
-    ];
+    const employerExcludeLower = listExtras.employerExclude.map((value) =>
+      value.toLowerCase(),
+    );
+    const settingsBlockedKeywords = listExtras.settingsBlockedEmployerKeywords;
     const uniqueEmployerExcludeLower = [...new Set(employerExcludeLower)];
 
     let filtered = [...jobs];
@@ -164,14 +166,24 @@ export const useFilteredJobs = (
 
     if (employerIncludeLower.length > 0) {
       filtered = filtered.filter((job) =>
-        employerMatchesAnyKeyword(job.employer, employerIncludeLower),
+        employerIncludeLower.some((keyword) =>
+          job.employer.toLowerCase().includes(keyword),
+        ),
+      );
+    }
+
+    if (settingsBlockedKeywords.length > 0) {
+      filtered = filtered.filter(
+        (job) => !jobMatchesAnyBlockedKeyword(job, settingsBlockedKeywords),
       );
     }
 
     if (uniqueEmployerExcludeLower.length > 0) {
       filtered = filtered.filter(
         (job) =>
-          !employerMatchesAnyKeyword(job.employer, uniqueEmployerExcludeLower),
+          !uniqueEmployerExcludeLower.some((keyword) =>
+            job.employer.toLowerCase().includes(keyword),
+          ),
       );
     }
 

orchestrator/src/client/pages/settings/components/ScoringSettingsSection.tsx

@@ -235,7 +235,7 @@ export const ScoringSettingsSection: React.FC<ScoringSettingsSectionProps> = ({
                 setValue("blockedCompanyKeywords", value, { shouldDirty: true })
               }
               placeholder='e.g. "recruitment", "staffing"'
-              helperText="Maintained here and saved with Settings. Each token is a case-insensitive substring match on the employer name. Matching jobs are dropped during discovery (not removed from the database if already imported). See docs: /docs/features/company-skip-list"
+              helperText="Maintained here and saved with Settings. Each token matches employer or job title (hyphens ignored, so co-op matches coop). Matching jobs are dropped during discovery. See docs: /docs/features/company-skip-list"
               removeLabelPrefix="Remove blocked keyword"
               disabled={isLoading || isSaving}
             />

orchestrator/src/server/pipeline/steps/discover-jobs.test.ts

@@ -394,6 +394,58 @@ describe("discoverJobsStep", () => {
     ]);
   });
 
+  it("drops co-op titles via company skip list and coop deal-breaker token", async () => {
+    const settingsRepo = await import("@server/repositories/settings");
+    const registryModule = await import("@server/extractors/registry");
+
+    const bcManifest = {
+      id: "bctenet",
+      displayName: "BC T-Net",
+      providesSources: ["bctenet"],
+      run: vi.fn().mockResolvedValue({
+        success: true,
+        jobs: [
+          {
+            source: "bctenet",
+            title: "Co-op Software Development Engineer in Test (SDET)",
+            employer: "Global Relay Communications Inc.",
+            location: "British Columbia, Canada",
+            jobUrl: "https://example.com/job-coop",
+          },
+          {
+            source: "bctenet",
+            title: "SDET",
+            employer: "Contoso",
+            location: "Vancouver, BC, Canada",
+            jobUrl: "https://example.com/job-sdet",
+          },
+        ],
+      }),
+    };
+
+    vi.mocked(settingsRepo.getAllSettings).mockResolvedValue({
+      searchTerms: JSON.stringify(["sdet"]),
+      searchCities: "Canada",
+      blockedCompanyKeywords: JSON.stringify(["co-op"]),
+    } as any);
+
+    vi.mocked(registryModule.getExtractorRegistry).mockResolvedValue({
+      manifests: new Map([["bctenet", bcManifest as any]]),
+      manifestBySource: new Map([["bctenet", bcManifest as any]]),
+      availableSources: ["bctenet"],
+    } as any);
+
+    const result = await discoverJobsStep({
+      mergedConfig: {
+        ...baseConfig,
+        sources: ["bctenet"],
+      },
+    });
+
+    expect(result.discoveredJobs).toHaveLength(1);
+    expect(result.discoveredJobs[0]?.jobUrl).toBe("https://example.com/job-sdet");
+  });
+
   it("drops jobs with blocked country in description when location is worldwide", async () => {
     const settingsRepo = await import("@server/repositories/settings");
     const registryModule = await import("@server/extractors/registry");

orchestrator/src/server/pipeline/steps/discover-jobs.ts

@@ -15,6 +15,7 @@ import {
   isSourceAllowedForCountry,
   normalizeCountryKey,
 } from "@shared/location-support.js";
+import { textMatchesAnyKeyword } from "@shared/keyword-match.js";
 import { resolveBlockedCompanyKeywordsFromStoredString } from "@shared/resolve-blocked-company-keywords.js";
 import {
   inferCountryKeyFromSearchGeography,
@@ -40,15 +41,14 @@ type DiscoverySourceTask = {
   run: () => Promise<DiscoveryTaskResult>;
 };
 
-function isBlockedEmployer(
-  employer: string | null | undefined,
-  blockedKeywordsLowerCase: string[],
+function jobMatchesBlockedCompanyKeywords(
+  job: CreateJobInput,
+  blockedKeywords: string[],
 ): boolean {
-  if (!employer) return false;
-  if (blockedKeywordsLowerCase.length === 0) return false;
-  const normalizedEmployer = employer.toLowerCase();
-  return blockedKeywordsLowerCase.some((keyword) =>
-    normalizedEmployer.includes(keyword),
+  if (blockedKeywords.length === 0) return false;
+  return (
+    textMatchesAnyKeyword(job.employer, blockedKeywords) ||
+    textMatchesAnyKeyword(job.title, blockedKeywords)
   );
 }
 
@@ -156,7 +156,6 @@ function filterJobsBySearchProfile(args: {
 
   const roleMatchers = buildRoleMatchers(targetRolePhrases);
   const mustHaveLower = mustHaveSkills.map(normalizeText).filter(Boolean);
-  const dealBreakersLower = dealBreakers.map(normalizeText).filter(Boolean);
 
   const filtered = jobs.filter((job) => {
     const title = normalizeText(job.title);
@@ -164,8 +163,8 @@ function filterJobsBySearchProfile(args: {
     const haystack = `${title}\n${body}`;
 
     if (
-      dealBreakersLower.length > 0 &&
-      matchesAny(haystack, dealBreakersLower)
+      dealBreakers.length > 0 &&
+      dealBreakers.some((breaker) => textMatchesAnyKeyword(haystack, breaker))
     ) {
       return false;
     }
@@ -521,11 +520,8 @@ export async function discoverJobsStep(args: {
   const blockedCompanyKeywords = resolveBlockedCompanyKeywordsFromStoredString(
     settings.blockedCompanyKeywords,
   );
-  const blockedKeywordsLowerCase = blockedCompanyKeywords.map((value) =>
-    value.toLowerCase(),
-  );
   const afterCompanyFilter = cityFilteredJobs.filter(
-    (job) => !isBlockedEmployer(job.employer, blockedKeywordsLowerCase),
+    (job) => !jobMatchesBlockedCompanyKeywords(job, blockedCompanyKeywords),
   );
   const companyDroppedCount =
     cityFilteredJobs.length - afterCompanyFilter.length;

orchestrator/src/server/services/scorer.test.ts

@@ -284,6 +284,49 @@ describe("salary penalty", () => {
     vi.restoreAllMocks();
   });
 
+  describe("profile deal-breakers", () => {
+    it("short-circuits scoring when co-op matches coop deal-breaker token", async () => {
+      const { scoreJobSuitability } = await import("./scorer");
+      const { LlmService } = await import("./llm/service");
+
+      const callJson = vi.spyOn(LlmService.prototype, "callJson");
+
+      getEffectiveSettingsMock.mockResolvedValue({
+        jobSearchProfile: {
+          value: {
+            targetRoles: ["SDET"],
+            experienceLevel: "Senior",
+            mustHaveSkills: [],
+            niceToHaveSkills: [],
+            dealBreakers: ["coop"],
+            preferredWorkArrangement: ["remote"],
+            preferredLocations: ["Canada"],
+            minimumSalary: "",
+            industriesToTarget: [],
+            industriesToAvoid: [],
+            aboutMe: "",
+          },
+          default: null,
+          override: null,
+        },
+        penalizeMissingSalary: { value: false, default: false, override: null },
+        missingSalaryPenalty: { value: 10, default: 10, override: null },
+        scoringInstructions: { value: "", default: "", override: null },
+        rxresumeBaseResumeId: "base-resume-123",
+      } as any);
+
+      const job = createJob({
+        title: "Co-op Software Development Engineer in Test (SDET)",
+        employer: "Global Relay Communications Inc.",
+      });
+      const result = await scoreJobSuitability(job, {});
+
+      expect(callJson).not.toHaveBeenCalled();
+      expect(result.score).toBeLessThanOrEqual(15);
+      expect(result.analysis?.dealBreakerHits?.length).toBeGreaterThan(0);
+    });
+  });
+
   describe("isSalaryMissing detection", () => {
     it("should detect null salary as missing", async () => {
       const { scoreJobSuitability } = await import("./scorer");

orchestrator/src/server/services/scorer.ts

@@ -3,6 +3,7 @@
  */
 
 import { logger } from "@infra/logger";
+import { textMatchesKeyword } from "@shared/keyword-match";
 import type { Job, JobSearchProfile, SuitabilityAnalysis } from "@shared/types";
 import { jobLikelyRequiresOfficePresence } from "@shared/work-arrangement";
 import { LlmService } from "./llm/service";
@@ -177,6 +178,30 @@ function extractAnalysis(data: ScoringLlmResponse): SuitabilityAnalysis | null {
  * Score a job's suitability based on profile and job description.
  * Includes retry logic for when AI returns garbage responses.
  */
+function scoreForProfileDealBreakers(
+  job: Job,
+  jobSearchProfile: JobSearchProfile,
+): SuitabilityResult | null {
+  if (jobSearchProfile.dealBreakers.length === 0) return null;
+  const haystack = `${job.title}\n${job.jobDescription ?? ""}`;
+  const hit = jobSearchProfile.dealBreakers.find((breaker) =>
+    textMatchesKeyword(haystack, breaker),
+  );
+  if (!hit) return null;
+  return {
+    score: 10,
+    reason: `Deal-breaker "${hit}" matched in the job title or description.`,
+    analysis: {
+      roleTypeMatch: 15,
+      workArrangementMatch: undefined,
+      strengths: [],
+      gaps: [`Posting matches profile deal-breaker: ${hit}`],
+      suggestions: ["Skip co-op, intern, and other excluded role types"],
+      dealBreakerHits: [`Profile deal-breaker: ${hit}`],
+    },
+  };
+}
+
 export async function scoreJobSuitability(
   job: Job,
   profile: Record<string, unknown>,
@@ -189,6 +214,11 @@ export async function scoreJobSuitability(
   const jobSearchProfile = settings.jobSearchProfile?.value ?? null;
   const hasProfile = jobSearchProfile && hasNonEmptyProfile(jobSearchProfile);
 
+  if (hasProfile && jobSearchProfile) {
+    const dealBreakerScore = scoreForProfileDealBreakers(job, jobSearchProfile);
+    if (dealBreakerScore) return dealBreakerScore;
+  }
+
   const prompt = buildScoringPrompt(job, sanitizeProfileForPrompt(profile), {
     instructions: settings.scoringInstructions?.value ?? "",
     jobSearchProfile: hasProfile ? jobSearchProfile : null,

shared/src/index.ts

@@ -3,5 +3,6 @@ export * from "./job-fingerprint";
 export * from "./job-url-canonical";
 export * from "./location-support";
 export * from "./work-arrangement";
+export * from "./keyword-match";
 export * from "./types";
 export * from "./utils/type-conversion";

shared/src/keyword-match.test.ts

@@ -0,0 +1,21 @@
+import { describe, expect, it } from "vitest";
+import { textMatchesAnyKeyword, textMatchesKeyword } from "./keyword-match.js";
+
+describe("keyword-match", () => {
+  it("matches co-op title against coop deal-breaker token", () => {
+    expect(
+      textMatchesKeyword(
+        "Co-op Software Development Engineer in Test (SDET)",
+        "coop",
+      ),
+    ).toBe(true);
+    expect(textMatchesKeyword("Co-op SDET", "co-op")).toBe(true);
+  });
+
+  it("matches intern in employer or title", () => {
+    expect(textMatchesAnyKeyword("Acme Intern Program", ["intern"])).toBe(true);
+    expect(
+      textMatchesAnyKeyword("Global Relay Communications Inc.", ["intern"]),
+    ).toBe(false);
+  });
+});

shared/src/keyword-match.ts

@@ -0,0 +1,27 @@
+/**
+ * Case-insensitive substring matching that ignores spaces, hyphens, and underscores
+ * so "co-op", "co op", and "coop" all match the same token.
+ */
+export function normalizeForKeywordMatch(value: string): string {
+  return value
+    .toLowerCase()
+    .replace(/[\s\-_]+/g, "")
+    .trim();
+}
+
+export function textMatchesKeyword(
+  haystack: string | null | undefined,
+  keyword: string,
+): boolean {
+  const normalizedHaystack = normalizeForKeywordMatch(haystack ?? "");
+  const normalizedKeyword = normalizeForKeywordMatch(keyword);
+  if (!normalizedKeyword) return false;
+  return normalizedHaystack.includes(normalizedKeyword);
+}
+
+export function textMatchesAnyKeyword(
+  haystack: string | null | undefined,
+  keywords: readonly string[],
+): boolean {
+  return keywords.some((keyword) => textMatchesKeyword(haystack, keyword));
+}