update default

Author: AlexandreYang

auto-improve-skills/cmd/skillbench/main.go

@@ -50,7 +50,7 @@ func main() {
 		caseTimeout              = flag.Duration("case-timeout", 6*time.Minute, "timeout per benchmark case")
 		judge                    = flag.Bool("judge", false, "run optional LLM-as-judge scoring pass")
 		judgeWeight              = flag.Float64("judge-weight", 0.3, "when -judge is set, final score weight for judge score (0..1)")
-		objectiveQualityWeight   = flag.Float64("objective-quality-weight", 0.85, "composite objective weight for answer quality")
+		objectiveQualityWeight   = flag.Float64("objective-quality-weight", 0.80, "composite objective weight for answer quality")
 		objectiveDurationWeight  = flag.Float64("objective-duration-weight", 0.10, "composite objective weight for wall-clock investigation duration")
 		objectiveSkillSizeWeight = flag.Float64("objective-skill-size-weight", 0.10, "composite objective weight for skill size")
 		durationBudget           = flag.Duration("duration-budget", 60*time.Second, "per-case wall-clock duration with no objective penalty")

auto-improve-skills/cmd/skilltrain/main_test.go

@@ -475,7 +475,7 @@ func TestFormatCommitBodyIncludesChangeAndScoreDetails(t *testing.T) {
 		SkillSizeBytes:             8400,
 		SkillSizeScore:             0.93,
 		ObjectiveConfig: autoresearch.ObjectiveConfig{
-			QualityWeight:            0.85,
+			QualityWeight:            0.80,
 			DurationWeight:           0.10,
 			SkillSizeWeight:          0.05,
 			DurationBudgetSeconds:    120,
@@ -513,7 +513,7 @@ func TestFormatCommitBodyIncludesChangeAndScoreDetails(t *testing.T) {
 		"Objective: 94.25/100.00 (93.02% -> 94.25%, delta +1.23 pp)",
 		"Average case duration: 82.3s",
 		"Skill size: 2100 estimated tokens, 8400 bytes",
-		"Objective config: quality=0.85 duration=0.10 skill_size=0.05",
+		"Objective config: quality=0.80 duration=0.10 skill_size=0.05",
 		"Researcher summary:",
 		"Tightened the workflow",
 		"Change summary:",