fix(evals): reject contradictory missing-skill recovery claims

Co-Authored-By: Paperclip <noreply@paperclip.ing>

tests/runner-e2e/chat-catalog-domains/external-team-conditional.test.ts

@@ -260,3 +260,11 @@ it("does not credit resolved or all-ready claims as an unresolved occupancy repo
     expect(truthfulTeamPartial(body, skill)).toBe(false);
   }
 });
+
+it("rejects a later same-skill current recovery claim while saved state still shows the required catalog skill missing", () => {
+  const o = valid();
+  o.phaseReplies[0]!.body = "The existing CTO was preserved and the imported role is CTO 2. Launch Checklist's slug is occupied. Launch Checklist is now installed under another slug.";
+  expect(gradeTeamConditions(o, "initial").passed).toBe(false);
+  o.phaseReplies[0]!.body = "The existing CTO was preserved and the imported role is CTO 2. Launch Checklist's slug is occupied. The other three required skills installed.";
+  expect(gradeTeamConditions(o, "initial").passed).toBe(true);
+});

tests/runner-e2e/chat-catalog-domains/external-team-conditional.ts

@@ -266,7 +266,10 @@ export function truthfulTeamPartial(
       )
     );
   });
-  return reported && !falseFull;
+  const subjects = [skill.slug, skill.slug.replaceAll("-", " "), skill.name]
+    .map(name => name.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")).join("|");
+  const currentRecovery = new RegExp("(?:\\b(?:" + subjects + ")[*_`]*|^(?:it|(?:this|that) skill))\\s+(?:skill\\s+)?(?:is|has been)\\s+(?:(?:now|already|currently)\\s+)?(?:installed|available|resolved|recovered)\\b", "i");
+  return reported && !falseFull && !namedTeamSkillClauses(reply, skill, true).some(clause => currentRecovery.test(clause));
 }
 export function gradeTeamConditions(
   o: ExternalObservation,