Hardened the proposal modules to skip malformed proposals rather than hard failing the entire run

cmd/audita/main_integration_test.go

@@ -364,7 +364,7 @@ func TestProcessSuccessLargeTranscriptSubprocess(t *testing.T) {
 	}
 }
 
-func TestProcessFailureMalformedStructuredLLMResponseViaSubprocessHook(t *testing.T) {
+func TestProcessMalformedStructuredLLMResponseViaSubprocessHookSoftSkips(t *testing.T) {
 	workDir := t.TempDir()
 	reportPath := filepath.Join(t.TempDir(), "report.json")
 	result := runCLISubprocessWithEnv(t,
@@ -382,25 +382,34 @@ func TestProcessFailureMalformedStructuredLLMResponseViaSubprocessHook(t *testin
 		"--work-dir-retention",
 		"always",
 	)
-	if result.exitCode == 0 {
+	if result.exitCode != 0 {
-		t.Fatalf("expected nonzero exit code")
+		t.Fatalf("expected success exit code, got %d stderr=%q", result.exitCode, result.stderr)
 	}
-	if result.stdout != "" {
+	if !json.Valid([]byte(result.stdout)) {
-		t.Fatalf("expected empty stdout on failure, got %q", result.stdout)
+		t.Fatalf("expected transcript JSON on stdout, got %q", result.stdout)
 	}
-	if !strings.Contains(result.stderr, "runner_execution") {
+	if result.stderr != "" {
-		t.Fatalf("expected runner_execution failure, got %q", result.stderr)
+		t.Fatalf("expected quiet stderr on success, got %q", result.stderr)
-	}
-	if !strings.Contains(result.stderr, "diagnostics:") {
-		t.Fatalf("expected diagnostics path in stderr, got %q", result.stderr)
 	}
 	report := readFile(t, reportPath)
 	if !json.Valid(report) {
-		t.Fatalf("expected valid failure report JSON")
+		t.Fatalf("expected valid report JSON")
 	}
-	runDir := onlyRunDir(t, workDir)
+	var parsed struct {
-	if _, err := os.Stat(filepath.Join(runDir, "error.log")); err != nil {
+		Status        string `json:"status"`
-		t.Fatalf("expected error.log, got: %v", err)
+		ModuleResults []struct {
+			Status            string `json:"status"`
+			ValidatorRejected []any  `json:"validator_rejected"`
+		} `json:"module_results"`
+	}
+	if err := json.Unmarshal(report, &parsed); err != nil {
+		t.Fatalf("unmarshal report: %v", err)
+	}
+	if parsed.Status != "success" {
+		t.Fatalf("expected success report status, got %+v", parsed)
+	}
+	if len(parsed.ModuleResults) == 0 || len(parsed.ModuleResults[0].ValidatorRejected) == 0 {
+		t.Fatalf("expected soft-skip validator rejection records, got %+v", parsed.ModuleResults)
 	}
 }
 
@@ -436,7 +445,7 @@ func TestProcessFailureBackendLLMViaSubprocessHook(t *testing.T) {
 	}
 }
 
-func TestProcessFailureMidPipelinePreservesPartialReportsSubprocess(t *testing.T) {
+func TestProcessMidPipelineMalformedCorrectionsSoftSkipSubprocess(t *testing.T) {
 	workDir := t.TempDir()
 	reportPath := filepath.Join(t.TempDir(), "report.json")
 	result := runCLISubprocessWithEnv(t,
@@ -454,11 +463,11 @@ func TestProcessFailureMidPipelinePreservesPartialReportsSubprocess(t *testing.T
 		"--work-dir-retention",
 		"always",
 	)
-	if result.exitCode == 0 {
+	if result.exitCode != 0 {
-		t.Fatalf("expected nonzero exit code")
+		t.Fatalf("expected success exit code, got %d stderr=%q", result.exitCode, result.stderr)
 	}
-	if result.stdout != "" {
+	if !json.Valid([]byte(result.stdout)) {
-		t.Fatalf("expected empty stdout on failure, got %q", result.stdout)
+		t.Fatalf("expected transcript JSON on stdout, got %q", result.stdout)
 	}
 	reportRaw := readFile(t, reportPath)
 	var report struct {
@@ -472,11 +481,11 @@ func TestProcessFailureMidPipelinePreservesPartialReportsSubprocess(t *testing.T
 	if err := json.Unmarshal(reportRaw, &report); err != nil {
 		t.Fatalf("unmarshal report: %v", err)
 	}
-	if report.Status != "failed" || report.ErrorPhase != "runner_execution" {
+	if report.Status != "success" || report.ErrorPhase != "" {
-		t.Fatalf("expected failed runner_execution report, got %+v", report)
+		t.Fatalf("expected successful report with no runner_execution phase, got %+v", report)
 	}
 	if len(report.ModuleResults) == 0 {
-		t.Fatalf("expected partial module results in failure report")
+		t.Fatalf("expected module results in report")
 	}
 }
 

internal/framework/contracts/contracts.go

@@ -101,6 +101,7 @@ type ProposalRequest struct {
 	RunSpec      ModuleRunSpec       `json:"run_spec"`
 	LLMClient    StructuredLLMClient `json:"-"`
 	LLMScheduler LLMScheduler        `json:"-"`
+	OnDroppedCandidate func(index int, targetSegmentID int, originalText string, correctedText string, confidence float64, reasonCode string, message string) `json:"-"`
 }
 
 // ValidationRequest is the input to validator execution.

internal/framework/proposal_generation/generate.go

@@ -44,6 +44,16 @@ type StructuredCorrectionSet struct {
 	Corrections []StructuredCorrectionProposal `json:"corrections"`
 }
 
+type DroppedCandidate struct {
+	ProposalIndex   int     `json:"proposal_index"`
+	TargetSegmentID int     `json:"target_segment_id"`
+	OriginalText    string  `json:"original_text"`
+	CorrectedText   string  `json:"corrected_text"`
+	Confidence      float64 `json:"confidence"`
+	ReasonCode      string  `json:"reason_code"`
+	Message         string  `json:"message"`
+}
+
 // Request captures reusable proposal-generation inputs for future modules.
 type Request struct {
 	ModuleKey         string                      `json:"module_key"`
@@ -62,12 +72,14 @@ type Request struct {
 	Scheduler         contracts.LLMScheduler
 	DiagnosticsDir    string
 	DiagnosticsWriter InteractionDiagnosticsWriter
+	OnDroppedCandidate func(dropped DroppedCandidate)
 }
 
 // Result contains generated candidate proposals and optional diagnostics paths.
 type Result struct {
 	Corrections []proposals.CorrectionProposal         `json:"corrections"`
 	Enriched    []proposals.EnrichedCorrectionProposal `json:"enriched"`
+	Dropped     []DroppedCandidate                     `json:"dropped,omitempty"`
 	Artifacts   InteractionArtifacts                   `json:"artifacts,omitempty"`
 }
 
@@ -161,7 +173,9 @@ func GenerateCandidates(ctx context.Context, req Request) (Result, error) {
 
 	corrections := make([]proposals.CorrectionProposal, 0, len(response.Corrections))
 	enriched := make([]proposals.EnrichedCorrectionProposal, 0, len(response.Corrections))
+	dropped := make([]DroppedCandidate, 0)
 	for i, raw := range response.Corrections {
+		proposalIndex := req.StartIndex + i
 		candidate := proposals.CorrectionProposal{
 			TargetSegmentID: raw.TargetSegmentID,
 			OriginalText:    raw.OriginalText,
@@ -169,14 +183,27 @@ func GenerateCandidates(ctx context.Context, req Request) (Result, error) {
 			Confidence:      raw.Confidence,
 		}
 		if err := candidate.Validate(); err != nil {
-			return Result{}, fmt.Errorf("invalid structured correction at index %d: %w", i, err)
+			record := DroppedCandidate{
+				ProposalIndex:   proposalIndex,
+				TargetSegmentID: raw.TargetSegmentID,
+				OriginalText:    raw.OriginalText,
+				CorrectedText:   raw.CorrectedText,
+				Confidence:      raw.Confidence,
+				ReasonCode:      "invalid_structured_correction",
+				Message:         fmt.Sprintf("invalid structured correction at index %d: %v", i, err),
+			}
+			dropped = append(dropped, record)
+			if req.OnDroppedCandidate != nil {
+				req.OnDroppedCandidate(record)
+			}
+			continue
 		}
 
 		corrections = append(corrections, candidate)
 		enrichedCandidate := proposals.EnrichedCorrectionProposal{
 			CorrectionProposal: candidate,
 			ProposalMetadata: proposals.ProposalMetadata{
-				ProposalIndex:  req.StartIndex + i,
+				ProposalIndex:  proposalIndex,
 				ModuleKey:      req.ModuleKey,
 				ModuleInstance: req.ModuleInstance,
 			},
@@ -191,6 +218,7 @@ func GenerateCandidates(ctx context.Context, req Request) (Result, error) {
 	return Result{
 		Corrections: corrections,
 		Enriched:    enriched,
+		Dropped:     dropped,
 		Artifacts:   artifacts,
 	}, nil
 }

internal/framework/proposal_generation/generate_test.go

@@ -229,7 +229,7 @@ func TestGenerateCandidatesMalformedStructuredResponse(t *testing.T) {
 		responses: []StructuredCorrectionSet{
 			{
 				Corrections: []StructuredCorrectionProposal{
-					{TargetSegmentID: 1, OriginalText: "x", CorrectedText: "", Confidence: 0.9},
+					{TargetSegmentID: 0, OriginalText: "x", CorrectedText: "", Confidence: 0.9},
 				},
 			},
 		},
@@ -237,9 +237,18 @@ func TestGenerateCandidatesMalformedStructuredResponse(t *testing.T) {
 	req := defaultRequest(t)
 	req.LLMClient = client
 
-	_, err := GenerateCandidates(context.Background(), req)
+	got, err := GenerateCandidates(context.Background(), req)
-	if err == nil || !strings.Contains(err.Error(), "invalid structured correction") {
+	if err != nil {
-		t.Fatalf("expected structured response validation failure, got %v", err)
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(got.Corrections) != 0 || len(got.Enriched) != 0 {
+		t.Fatalf("expected invalid proposal to be dropped, got %+v", got)
+	}
+	if len(got.Dropped) != 1 {
+		t.Fatalf("expected one dropped candidate, got %+v", got.Dropped)
+	}
+	if got.Dropped[0].ReasonCode != "invalid_structured_correction" {
+		t.Fatalf("unexpected dropped reason: %+v", got.Dropped[0])
 	}
 }
 

internal/framework/proposals/proposal.go

@@ -38,9 +38,6 @@ func (p CorrectionProposal) Validate() error {
 	if strings.TrimSpace(p.OriginalText) == "" {
 		return fmt.Errorf("proposal original_text must not be empty")
 	}
-	if strings.TrimSpace(p.CorrectedText) == "" {
-		return fmt.Errorf("proposal corrected_text must not be empty")
-	}
 	if p.Confidence < 0.0 || p.Confidence > 1.0 {
 		return fmt.Errorf("proposal confidence must be between 0.0 and 1.0")
 	}

internal/framework/proposals/proposal_test.go

@@ -35,7 +35,7 @@ func TestCorrectionProposalValidate_InvalidEmptyOriginalText(t *testing.T) {
 	}
 }
 
-func TestCorrectionProposalValidate_InvalidEmptyCorrectedText(t *testing.T) {
+func TestCorrectionProposalValidate_AllowsEmptyCorrectedText(t *testing.T) {
 	proposal := CorrectionProposal{
 		TargetSegmentID: 42,
 		OriginalText:    "gestures",
@@ -44,11 +44,8 @@ func TestCorrectionProposalValidate_InvalidEmptyCorrectedText(t *testing.T) {
 	}
 
 	err := proposal.Validate()
-	if err == nil {
+	if err != nil {
-		t.Fatal("expected validation error, got nil")
+		t.Fatalf("expected valid proposal, got %v", err)
-	}
-	if err.Error() != "proposal corrected_text must not be empty" {
-		t.Fatalf("unexpected error: %v", err)
 	}
 }
 

internal/framework/runner/runner.go

@@ -238,6 +238,8 @@ type sectionProposalResult struct {
 	sectionPos int
 	section    chunking.Section
 	corrected  []proposals.CorrectionProposal
+	dropped    []droppedProposalCandidate
+	rawCount   int
 	err        error
 }
 
@@ -257,6 +259,22 @@ type modulePipelineResult struct {
 	ValidatorRejected  []ValidatorRejectedChange
 }
 
+type droppedProposalCandidate struct {
+	ProposalIndex   int
+	TargetSegmentID int
+	OriginalText    string
+	CorrectedText   string
+	ReasonCode      string
+	Message         string
+}
+
+const (
+	validatorNameProposalGeneration  = "proposal_generation"
+	reasonValidatorExecutionError    = "validator_execution_error"
+	reasonValidatorMalformedResponse = "validator_malformed_response"
+	reasonValidatorMissingDecision   = "validator_missing_decision"
+)
+
 func collectSectionProposals(ctx context.Context, input collectSectionProposalsInput) (context.Context, <-chan sectionProposalResult, context.CancelFunc) {
 	results := make(chan sectionProposalResult, len(input.Sections))
 	runCtx, cancel := context.WithCancel(ctx)
@@ -271,6 +289,7 @@ func collectSectionProposals(ctx context.Context, input collectSectionProposalsI
 			go func() {
 				defer wg.Done()
 				meta := contracts.SectionMetadataFromSection(section)
+				dropped := make([]droppedProposalCandidate, 0)
 				corrected, err := input.Module.Propose(withModuleInstanceContext(runCtx, input.Spec.InstanceName), contracts.ProposalRequest{
 					ExecutionContext: contracts.ExecutionContext{
 						Config:            input.Config,
@@ -286,12 +305,25 @@ func collectSectionProposals(ctx context.Context, input collectSectionProposalsI
 					},
 					LLMClient:    input.ProposalClient,
 					LLMScheduler: input.ProposalScheduler,
+					OnDroppedCandidate: func(index int, targetSegmentID int, originalText string, correctedText string, confidence float64, reasonCode string, message string) {
+						_ = confidence
+						dropped = append(dropped, droppedProposalCandidate{
+							ProposalIndex:   index,
+							TargetSegmentID: targetSegmentID,
+							OriginalText:    originalText,
+							CorrectedText:   correctedText,
+							ReasonCode:      reasonCode,
+							Message:         message,
+						})
+					},
 				})
 				select {
 				case results <- sectionProposalResult{
 					sectionPos: sectionPos,
 					section:    section,
 					corrected:  corrected,
+					dropped:    dropped,
+					rawCount:   len(corrected) + len(dropped),
 					err:        err,
 				}:
 				case <-runCtx.Done():
@@ -362,7 +394,7 @@ func runModulePipeline(ctx context.Context, input collectSectionProposalsInput)
 			delete(pending, nextSectionToProcess)
 			sectionMeta := contracts.SectionMetadataFromSection(sectionResult.section)
 			sectionEnriched := make([]proposals.EnrichedCorrectionProposal, 0, len(sectionResult.corrected))
-			for i, corrected := range sectionResult.corrected {
+				for i, corrected := range sectionResult.corrected {
 				sectionIndex := sectionMeta.Index
 				sectionEnriched = append(sectionEnriched, proposals.EnrichedCorrectionProposal{
 					CorrectionProposal: corrected,
@@ -373,9 +405,22 @@ func runModulePipeline(ctx context.Context, input collectSectionProposalsInput)
 						SectionIndex:   &sectionIndex,
 					},
 				})
-			}
+				}
-			nextProposalIndex += len(sectionEnriched)
+				out.ProposalCount += sectionResult.rawCount
-			out.ProposalCount += len(sectionEnriched)
+				nextProposalIndex += sectionResult.rawCount
+				for _, dropped := range sectionResult.dropped {
+					out.ValidatorRejected = append(out.ValidatorRejected, ValidatorRejectedChange{
+						ValidatorName:   validatorNameProposalGeneration,
+						ProposalIndex:   dropped.ProposalIndex,
+						ModuleKey:       input.Spec.ModuleKey,
+						ModuleInstance:  input.Spec.InstanceName,
+						TargetSegmentID: dropped.TargetSegmentID,
+						OriginalText:    dropped.OriginalText,
+						CorrectedText:   dropped.CorrectedText,
+						ReasonCode:      dropped.ReasonCode,
+						Message:         dropped.Message,
+					})
+				}
 
 			validationLaunches++
 			vwg.Add(1)
@@ -508,54 +553,33 @@ func validateSectionCandidates(ctx context.Context, input validateSectionCandida
 			)
 		}
 		if err != nil {
-			return validateSectionCandidatesResult{
+			for _, p := range eligible {
-				approved:  eligible,
+				decisions = append(decisions, ValidatorDecisionRecord{
-				decisions: decisions,
+					ValidatorName: validator.Name(),
-				rejected:  rejected,
+					ProposalIndex: p.ProposalIndex,
-			}, fmt.Errorf("validator %q failed: %w", validator.Name(), err)
+					Approved:      false,
-		}
+					ReasonCode:    reasonValidatorExecutionError,
-		if err := validators.EnforceDecisionCardinality(eligible, vResult.Decisions); err != nil {
+					Message:       fmt.Sprintf("validator execution failed: %v", err),
-			return validateSectionCandidatesResult{
+				})
-				approved:  eligible,
+				rejected = append(rejected, ValidatorRejectedChange{
-				decisions: decisions,
+					ValidatorName:   validator.Name(),
-				rejected:  rejected,
+					ProposalIndex:   p.ProposalIndex,
-			}, fmt.Errorf("validator %q cardinality failed: %w", validator.Name(), err)
+					ModuleKey:       p.ModuleKey,
-		}
+					ModuleInstance:  p.ModuleInstance,
-
+					TargetSegmentID: p.TargetSegmentID,
-		nextEligible := make([]proposals.EnrichedCorrectionProposal, 0, len(eligible))
+					OriginalText:    p.OriginalText,
-		byIndex := make(map[int]proposals.EnrichedCorrectionProposal, len(eligible))
+					CorrectedText:   p.CorrectedText,
-		for _, p := range eligible {
+					ReasonCode:      reasonValidatorExecutionError,
-			byIndex[p.ProposalIndex] = p
+					Message:         fmt.Sprintf("validator execution failed: %v", err),
-		}
+				})
-
-		for _, d := range vResult.Decisions {
-			decisions = append(decisions, ValidatorDecisionRecord{
-				ValidatorName:          validator.Name(),
-				ProposalIndex:          d.ProposalIndex,
-				Approved:               d.Approved,
-				ReasonCode:             d.ReasonCode,
-				Message:                d.Message,
-				DiagnosticArtifactPath: d.DiagnosticArtifactPath,
-			})
-			if d.Approved {
-				nextEligible = append(nextEligible, byIndex[d.ProposalIndex])
-				continue
 			}
-			p := byIndex[d.ProposalIndex]
+			eligible = nil
-			rejected = append(rejected, ValidatorRejectedChange{
+			continue
-				ValidatorName:   validator.Name(),
-				ProposalIndex:   p.ProposalIndex,
-				ModuleKey:       p.ModuleKey,
-				ModuleInstance:  p.ModuleInstance,
-				TargetSegmentID: p.TargetSegmentID,
-				OriginalText:    p.OriginalText,
-				CorrectedText:   p.CorrectedText,
-				ReasonCode:      d.ReasonCode,
-				Message:         d.Message,
-			})
 		}
-
+		reconciled := reconcileValidatorDecisions(validator.Name(), eligible, vResult.Decisions)
-		eligible = nextEligible
+		decisions = append(decisions, reconciled.decisions...)
+		rejected = append(rejected, reconciled.rejected...)
+		eligible = reconciled.nextEligible
 	}
 
 	return validateSectionCandidatesResult{
@@ -565,6 +589,114 @@ func validateSectionCandidates(ctx context.Context, input validateSectionCandida
 	}, nil
 }
 
+type reconciledValidatorDecisions struct {
+	nextEligible []proposals.EnrichedCorrectionProposal
+	decisions    []ValidatorDecisionRecord
+	rejected     []ValidatorRejectedChange
+}
+
+func reconcileValidatorDecisions(validatorName string, eligible []proposals.EnrichedCorrectionProposal, in []validators.Decision) reconciledValidatorDecisions {
+	byIndex := make(map[int]proposals.EnrichedCorrectionProposal, len(eligible))
+	for _, p := range eligible {
+		byIndex[p.ProposalIndex] = p
+	}
+
+	seen := make(map[int]struct{}, len(in))
+	forcedReject := make(map[int]bool)
+	valid := make(map[int]validators.Decision, len(in))
+	for _, d := range in {
+		if _, ok := byIndex[d.ProposalIndex]; !ok {
+			continue
+		}
+		if _, dup := seen[d.ProposalIndex]; dup {
+			forcedReject[d.ProposalIndex] = true
+			continue
+		}
+		seen[d.ProposalIndex] = struct{}{}
+		valid[d.ProposalIndex] = d
+	}
+
+	nextEligible := make([]proposals.EnrichedCorrectionProposal, 0, len(eligible))
+	decisions := make([]ValidatorDecisionRecord, 0, len(eligible))
+	rejected := make([]ValidatorRejectedChange, 0, len(eligible))
+
+	for _, p := range eligible {
+		idx := p.ProposalIndex
+		if forcedReject[idx] {
+			decisions = append(decisions, ValidatorDecisionRecord{
+				ValidatorName: validatorName,
+				ProposalIndex: idx,
+				Approved:      false,
+				ReasonCode:    reasonValidatorMalformedResponse,
+				Message:       "validator returned malformed response for this proposal index",
+			})
+			rejected = append(rejected, ValidatorRejectedChange{
+				ValidatorName:   validatorName,
+				ProposalIndex:   p.ProposalIndex,
+				ModuleKey:       p.ModuleKey,
+				ModuleInstance:  p.ModuleInstance,
+				TargetSegmentID: p.TargetSegmentID,
+				OriginalText:    p.OriginalText,
+				CorrectedText:   p.CorrectedText,
+				ReasonCode:      reasonValidatorMalformedResponse,
+				Message:         "validator returned malformed response for this proposal index",
+			})
+			continue
+		}
+		d, ok := valid[idx]
+		if !ok {
+			decisions = append(decisions, ValidatorDecisionRecord{
+				ValidatorName: validatorName,
+				ProposalIndex: idx,
+				Approved:      false,
+				ReasonCode:    reasonValidatorMissingDecision,
+				Message:       "validator did not return a decision for this proposal index",
+			})
+			rejected = append(rejected, ValidatorRejectedChange{
+				ValidatorName:   validatorName,
+				ProposalIndex:   p.ProposalIndex,
+				ModuleKey:       p.ModuleKey,
+				ModuleInstance:  p.ModuleInstance,
+				TargetSegmentID: p.TargetSegmentID,
+				OriginalText:    p.OriginalText,
+				CorrectedText:   p.CorrectedText,
+				ReasonCode:      reasonValidatorMissingDecision,
+				Message:         "validator did not return a decision for this proposal index",
+			})
+			continue
+		}
+		decisions = append(decisions, ValidatorDecisionRecord{
+			ValidatorName:          validatorName,
+			ProposalIndex:          d.ProposalIndex,
+			Approved:               d.Approved,
+			ReasonCode:             d.ReasonCode,
+			Message:                d.Message,
+			DiagnosticArtifactPath: d.DiagnosticArtifactPath,
+		})
+		if d.Approved {
+			nextEligible = append(nextEligible, p)
+			continue
+		}
+		rejected = append(rejected, ValidatorRejectedChange{
+			ValidatorName:   validatorName,
+			ProposalIndex:   p.ProposalIndex,
+			ModuleKey:       p.ModuleKey,
+			ModuleInstance:  p.ModuleInstance,
+			TargetSegmentID: p.TargetSegmentID,
+			OriginalText:    p.OriginalText,
+			CorrectedText:   p.CorrectedText,
+			ReasonCode:      d.ReasonCode,
+			Message:         d.Message,
+		})
+	}
+
+	return reconciledValidatorDecisions{
+		nextEligible: nextEligible,
+		decisions:    decisions,
+		rejected:     rejected,
+	}
+}
+
 func reorderValidatorsForPipeline(in []contracts.Validator) ([]contracts.Validator, map[string]int) {
 	deterministic := make([]contracts.Validator, 0, len(in))
 	llmBacked := make([]contracts.Validator, 0, len(in))

internal/framework/runner/runner_test.go

@@ -1054,7 +1054,7 @@ func TestRunnerMultipleValidatorsRunInOrderAndFilterSurvivors(t *testing.T) {
 	}
 }
 
-func TestRunnerValidatorCardinalityErrorStopsPipelineWithPartialProgress(t *testing.T) {
+func TestRunnerValidatorCardinalityErrorSoftRejectsAndContinues(t *testing.T) {
 	transcript := &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "teh cat"}}}
 	good := fakeModule{key: "m1", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
 		return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
@@ -1069,14 +1069,17 @@ func TestRunnerValidatorCardinalityErrorStopsPipelineWithPartialProgress(t *test
 	r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{"m1": good, "m2": bad}})
 
 	out, err := r.Run(context.Background(), RunInput{Transcript: transcript, ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "m1", InstanceName: "m1"}, {ModuleKey: "m2", InstanceName: "m2"}}})
-	if err == nil {
+	if err != nil {
-		t.Fatal("expected cardinality error")
+		t.Fatalf("unexpected run error: %v", err)
 	}
 	if out.FinalTranscript.Segments[0].Text != "the cat" {
-		t.Fatalf("expected partial progress preserved, got %q", out.FinalTranscript.Segments[0].Text)
+		t.Fatalf("expected second-module proposal to be rejected, got %q", out.FinalTranscript.Segments[0].Text)
 	}
-	if len(out.ModuleResults) != 2 || out.ModuleResults[1].Status != ModuleStatusFailed {
+	if len(out.ModuleResults) != 2 || out.ModuleResults[1].Status != ModuleStatusSuccess {
-		t.Fatalf("expected second module failed")
+		t.Fatalf("expected second module success with rejection, got %+v", out.ModuleResults)
+	}
+	if len(out.ModuleResults[1].ValidatorRejected) == 0 {
+		t.Fatalf("expected validator rejection record, got %+v", out.ModuleResults[1])
 	}
 }
 
@@ -1191,7 +1194,7 @@ func TestRunnerLLMValidatorRejectionPreventsApplication(t *testing.T) {
 	}
 }
 
-func TestRunnerLLMValidatorMalformedResponseFailsWithPartialProgress(t *testing.T) {
+func TestRunnerLLMValidatorMalformedResponseSoftRejectsWithProgress(t *testing.T) {
 	client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 99, Approved: true, Confidence: 0.9, Reason: "bad index"}}}}}
 	llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility, "")
 	r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
@@ -1209,15 +1212,21 @@ func TestRunnerLLMValidatorMalformedResponseFailsWithPartialProgress(t *testing.
 		ModuleSpecs:         []contracts.ModuleRunSpec{{ModuleKey: "m1", InstanceName: "m1"}, {ModuleKey: "m2", InstanceName: "m2"}},
 		ValidationLLMClient: client,
 	})
-	if err == nil {
+	if err != nil {
-		t.Fatal("expected llm validator failure")
+		t.Fatalf("unexpected run error: %v", err)
 	}
 	if out.FinalTranscript.Segments[0].Text != "the cat" {
 		t.Fatalf("expected partial progress retained")
 	}
+	if len(out.ModuleResults) != 2 || out.ModuleResults[1].Status != ModuleStatusSuccess {
+		t.Fatalf("expected second module success, got %+v", out.ModuleResults)
+	}
+	if len(out.ModuleResults[1].ValidatorRejected) == 0 {
+		t.Fatalf("expected llm validator rejection record")
+	}
 }
 
-func TestRunnerLLMValidatorMissingDecisionFails(t *testing.T) {
+func TestRunnerLLMValidatorMissingDecisionSoftRejects(t *testing.T) {
 	client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{}}}}
 	llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility, "")
 	r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
@@ -1226,18 +1235,21 @@ func TestRunnerLLMValidatorMissingDecisionFails(t *testing.T) {
 		}},
 	}})
 	cfg := config.Default()
-	_, err := r.Run(context.Background(), RunInput{
+	out, err := r.Run(context.Background(), RunInput{
 		Config:              &cfg,
 		Transcript:          &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "There were gestures at the temple."}}},
 		ModuleSpecs:         []contracts.ModuleRunSpec{{ModuleKey: "m", InstanceName: "m"}},
 		ValidationLLMClient: client,
 	})
-	if err == nil {
+	if err != nil {
-		t.Fatal("expected missing decision failure")
+		t.Fatalf("unexpected run error: %v", err)
+	}
+	if len(out.ModuleResults[0].ValidatorRejected) != 1 {
+		t.Fatalf("expected soft rejection, got %+v", out.ModuleResults[0].ValidatorRejected)
 	}
 }
 
-func TestRunnerLLMValidatorDuplicateDecisionFails(t *testing.T) {
+func TestRunnerLLMValidatorDuplicateDecisionSoftRejects(t *testing.T) {
 	client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{
 		{CorrectionIndex: 0, Approved: true, Confidence: 0.9, Reason: "ok"},
 		{CorrectionIndex: 0, Approved: false, Confidence: 0.9, Reason: "dup"},
@@ -1249,14 +1261,17 @@ func TestRunnerLLMValidatorDuplicateDecisionFails(t *testing.T) {
 		}},
 	}})
 	cfg := config.Default()
-	_, err := r.Run(context.Background(), RunInput{
+	out, err := r.Run(context.Background(), RunInput{
 		Config:              &cfg,
 		Transcript:          &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "There were gestures at the temple."}}},
 		ModuleSpecs:         []contracts.ModuleRunSpec{{ModuleKey: "m", InstanceName: "m"}},
 		ValidationLLMClient: client,
 	})
-	if err == nil {
+	if err != nil {
-		t.Fatal("expected duplicate decision failure")
+		t.Fatalf("unexpected run error: %v", err)
+	}
+	if len(out.ModuleResults[0].ValidatorRejected) != 1 {
+		t.Fatalf("expected soft rejection, got %+v", out.ModuleResults[0].ValidatorRejected)
 	}
 }
 

internal/framework/validators/deterministic.go

@@ -4,6 +4,9 @@ import (
 	"context"
 	"fmt"
 	"strings"
+
+	"gitea.maximumdirect.net/eric/audita/internal/core/schema"
+	"gitea.maximumdirect.net/eric/audita/internal/framework/proposals"
 )
 
 type ConfidenceThresholdValidator struct{}
@@ -62,11 +65,26 @@ type NonEmptyCorrectionValidator struct{}
 func (v NonEmptyCorrectionValidator) Name() string { return "non_empty_corrected_text" }
 
 func (v NonEmptyCorrectionValidator) Validate(_ context.Context, req Request) (Result, error) {
+	segments := make(map[int]string)
+	if req.WorkingTranscript != nil {
+		for _, seg := range req.WorkingTranscript.Segments {
+			segments[seg.ID] = seg.Text
+		}
+	}
+
 	decisions := make([]Decision, 0, len(req.CandidateProposal))
 	for _, c := range req.CandidateProposal {
-		if strings.TrimSpace(c.CorrectedText) == "" {
+		segmentText, ok := segments[c.TargetSegmentID]
-			decisions = append(decisions, rejection(c.ProposalIndex, ReasonEmptyCorrectedText, "corrected_text must not be empty"))
+		if ok {
-			continue
+			preview := proposals.PreviewProposalForSegment(
+				&schema.Segment{ID: c.TargetSegmentID, Text: segmentText},
+				c.CorrectionProposal,
+				req.ReplacementPolicy,
+			)
+			if preview.Applicable && strings.TrimSpace(preview.CorrectedSegmentText) == "" {
+				decisions = append(decisions, rejection(c.ProposalIndex, ReasonEmptyCorrectedText, "corrected segment text must not be empty"))
+				continue
+			}
 		}
 		decisions = append(decisions, approval(c.ProposalIndex))
 	}

internal/framework/validators/llm_validators.go

@@ -143,10 +143,7 @@ func (v *LLMBackedValidator) Validate(ctx context.Context, req Request) (Result,
 			return Result{}, fmt.Errorf("LLM validator %q completion failed: %w", v.name, err)
 		}
 
-		batchDecisions, err := mapLLMResponseToDecisions(batch.Items, response)
+		batchDecisions := mapLLMResponseToDecisions(batch.Items, response)
-		if err != nil {
-			return Result{}, fmt.Errorf("LLM validator %q response invalid: %w", v.name, err)
-		}
 		for i := range batchDecisions {
 			batchDecisions[i].DiagnosticArtifactPath = artifacts.ResponsePayloadPath
 		}
@@ -259,34 +256,48 @@ func promptBuilderForType(validatorType LLMValidatorType) (LLMPromptBuilder, err
 	}
 }
 
-func mapLLMResponseToDecisions(items []LLMValidationItem, response LLMValidationResponse) ([]Decision, error) {
+func mapLLMResponseToDecisions(items []LLMValidationItem, response LLMValidationResponse) []Decision {
 	expected := make(map[int]LLMValidationItem, len(items))
 	for _, item := range items {
 		expected[item.CorrectionIndex] = item
 	}
-	if len(response.Validations) == 0 {
-		return nil, fmt.Errorf("missing validations in structured response")
-	}
-
 	seen := make(map[int]LLMValidationDecision, len(response.Validations))
+	forcedReject := make(map[int]bool)
 	for _, d := range response.Validations {
-		if d.Confidence < 0.0 || d.Confidence > 1.0 {
-			return nil, fmt.Errorf("confidence for correction_index %d must be between 0.0 and 1.0", d.CorrectionIndex)
-		}
 		if _, ok := expected[d.CorrectionIndex]; !ok {
-			return nil, fmt.Errorf("unknown correction_index %d", d.CorrectionIndex)
+			continue
 		}
 		if _, exists := seen[d.CorrectionIndex]; exists {
-			return nil, fmt.Errorf("duplicate correction_index %d", d.CorrectionIndex)
+			forcedReject[d.CorrectionIndex] = true
+			continue
+		}
+		if d.Confidence < 0.0 || d.Confidence > 1.0 {
+			forcedReject[d.CorrectionIndex] = true
+			continue
 		}
 		seen[d.CorrectionIndex] = d
 	}
 
 	decisions := make([]Decision, 0, len(items))
 	for _, item := range items {
+		if forcedReject[item.CorrectionIndex] {
+			decisions = append(decisions, Decision{
+				ProposalIndex: item.CorrectionIndex,
+				Approved:      false,
+				ReasonCode:    ReasonValidatorMalformed,
+				Message:       "validator returned malformed decision payload for this proposal index",
+			})
+			continue
+		}
 		d, ok := seen[item.CorrectionIndex]
 		if !ok {
-			return nil, fmt.Errorf("missing correction_index %d", item.CorrectionIndex)
+			decisions = append(decisions, Decision{
+				ProposalIndex: item.CorrectionIndex,
+				Approved:      false,
+				ReasonCode:    ReasonValidatorMissing,
+				Message:       "validator did not return a decision for this proposal index",
+			})
+			continue
 		}
 		reasonCode := ReasonApproved
 		if !d.Approved {
@@ -299,5 +310,5 @@ func mapLLMResponseToDecisions(items []LLMValidationItem, response LLMValidation
 			Message:       strings.TrimSpace(d.Reason),
 		})
 	}
-	return decisions, nil
+	return decisions
 }

internal/framework/validators/llm_validators_test.go

@@ -259,18 +259,24 @@ func TestLLMBackedValidatorMalformedOutputFails(t *testing.T) {
 	}
 }
 
-func TestLLMBackedValidatorMissingDecisionFails(t *testing.T) {
+func TestLLMBackedValidatorMissingDecisionSoftRejects(t *testing.T) {
 	client := &fakeStructuredLLMClient{responses: []LLMValidationResponse{{Validations: []LLMValidationDecision{}}}}
 	v, _ := NewLLMBackedValidator("spoken_form_plausibility_review", LLMValidatorTypeSpokenFormPlausibility, "test-model")
 	req := makeReq([]proposals.EnrichedCorrectionProposal{mk(0, "gestures", "Jesters")})
 	req.LLMClient = client
-	_, err := v.Validate(context.Background(), req)
+	res, err := v.Validate(context.Background(), req)
-	if err == nil || !strings.Contains(err.Error(), "response invalid") {
+	if err != nil {
-		t.Fatalf("expected missing decision error, got %v", err)
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(res.Decisions) != 1 || res.Decisions[0].Approved {
+		t.Fatalf("expected one soft rejection, got %+v", res.Decisions)
+	}
+	if res.Decisions[0].ReasonCode != ReasonValidatorMissing {
+		t.Fatalf("expected %q, got %+v", ReasonValidatorMissing, res.Decisions[0])
 	}
 }
 
-func TestLLMBackedValidatorDuplicateDecisionFails(t *testing.T) {
+func TestLLMBackedValidatorDuplicateDecisionSoftRejects(t *testing.T) {
 	client := &fakeStructuredLLMClient{responses: []LLMValidationResponse{{Validations: []LLMValidationDecision{
 		{CorrectionIndex: 0, Approved: true, Confidence: 0.9, Reason: "ok"},
 		{CorrectionIndex: 0, Approved: false, Confidence: 0.9, Reason: "dup"},
@@ -278,20 +284,32 @@ func TestLLMBackedValidatorDuplicateDecisionFails(t *testing.T) {
 	v, _ := NewLLMBackedValidator("spoken_form_plausibility_review", LLMValidatorTypeSpokenFormPlausibility, "test-model")
 	req := makeReq([]proposals.EnrichedCorrectionProposal{mk(0, "gestures", "Jesters")})
 	req.LLMClient = client
-	_, err := v.Validate(context.Background(), req)
+	res, err := v.Validate(context.Background(), req)
-	if err == nil || !strings.Contains(err.Error(), "duplicate") {
+	if err != nil {
-		t.Fatalf("expected duplicate decision error, got %v", err)
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(res.Decisions) != 1 || res.Decisions[0].Approved {
+		t.Fatalf("expected one soft rejection, got %+v", res.Decisions)
+	}
+	if res.Decisions[0].ReasonCode != ReasonValidatorMalformed {
+		t.Fatalf("expected %q, got %+v", ReasonValidatorMalformed, res.Decisions[0])
 	}
 }
 
-func TestLLMBackedValidatorUnknownProposalIndexFails(t *testing.T) {
+func TestLLMBackedValidatorUnknownProposalIndexSoftRejectsMissing(t *testing.T) {
 	client := &fakeStructuredLLMClient{responses: []LLMValidationResponse{{Validations: []LLMValidationDecision{{CorrectionIndex: 99, Approved: true, Confidence: 0.9, Reason: "unknown"}}}}}
 	v, _ := NewLLMBackedValidator("spoken_form_plausibility_review", LLMValidatorTypeSpokenFormPlausibility, "test-model")
 	req := makeReq([]proposals.EnrichedCorrectionProposal{mk(0, "gestures", "Jesters")})
 	req.LLMClient = client
-	_, err := v.Validate(context.Background(), req)
+	res, err := v.Validate(context.Background(), req)
-	if err == nil || !strings.Contains(err.Error(), "unknown") {
+	if err != nil {
-		t.Fatalf("expected unknown index error, got %v", err)
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(res.Decisions) != 1 || res.Decisions[0].Approved {
+		t.Fatalf("expected one soft rejection, got %+v", res.Decisions)
+	}
+	if res.Decisions[0].ReasonCode != ReasonValidatorMissing {
+		t.Fatalf("expected %q, got %+v", ReasonValidatorMissing, res.Decisions[0])
 	}
 }
 

internal/framework/validators/models.go

@@ -18,6 +18,8 @@ const (
 	ReasonEmptyCorrectedText    = "empty_corrected_text"
 	ReasonNoEffect              = "no_effect"
 	ReasonProtectedGlossaryTerm = "protected_glossary_term"
+	ReasonValidatorMalformed    = "validator_malformed_response"
+	ReasonValidatorMissing      = "validator_missing_decision"
 )
 
 // Request is the runtime input shared by deterministic validators.

internal/framework/validators/validators_test.go

@@ -90,10 +90,14 @@ func TestOriginalTextPresenceValidator(t *testing.T) {
 }
 
 func TestNonEmptyCorrectionValidator(t *testing.T) {
-	req := Request{CandidateProposal: []proposals.EnrichedCorrectionProposal{
+	req := Request{
+		WorkingTranscript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "hello"}}},
+		ReplacementPolicy: proposals.ReplacementPolicyRequireUnique,
+		CandidateProposal: []proposals.EnrichedCorrectionProposal{
 		mkCandidate(0, 1, "hello", "hi", 0.9),
 		mkCandidate(1, 1, "hello", "   ", 0.9),
-	}}
+		},
+	}
 	res, err := (NonEmptyCorrectionValidator{}).Validate(context.Background(), req)
 	if err != nil {
 		t.Fatalf("Validate error: %v", err)
@@ -154,6 +158,8 @@ func TestStableReasonCodes(t *testing.T) {
 		ReasonEmptyCorrectedText,
 		ReasonNoEffect,
 		ReasonProtectedGlossaryTerm,
+		ReasonValidatorMalformed,
+		ReasonValidatorMissing,
 	}
 	for _, code := range codes {
 		if strings.TrimSpace(code) == "" {

internal/modules/glossary/module.go

@@ -72,6 +72,20 @@ func (m *Module) Propose(ctx context.Context, req contracts.ProposalRequest) ([]
 		LLMClient:      req.LLMClient,
 		Scheduler:      req.LLMScheduler,
 		DiagnosticsDir: req.DiagnosticsDir,
+		OnDroppedCandidate: func(dropped proposal_generation.DroppedCandidate) {
+			if req.OnDroppedCandidate == nil {
+				return
+			}
+			req.OnDroppedCandidate(
+				dropped.ProposalIndex,
+				dropped.TargetSegmentID,
+				dropped.OriginalText,
+				dropped.CorrectedText,
+				dropped.Confidence,
+				dropped.ReasonCode,
+				dropped.Message,
+			)
+		},
 	})
 	if err != nil {
 		return nil, err

internal/modules/grammar/module.go

@@ -72,6 +72,20 @@ func (m *Module) Propose(ctx context.Context, req contracts.ProposalRequest) ([]
 		LLMClient:      req.LLMClient,
 		Scheduler:      req.LLMScheduler,
 		DiagnosticsDir: req.DiagnosticsDir,
+		OnDroppedCandidate: func(dropped proposal_generation.DroppedCandidate) {
+			if req.OnDroppedCandidate == nil {
+				return
+			}
+			req.OnDroppedCandidate(
+				dropped.ProposalIndex,
+				dropped.TargetSegmentID,
+				dropped.OriginalText,
+				dropped.CorrectedText,
+				dropped.Confidence,
+				dropped.ReasonCode,
+				dropped.Message,
+			)
+		},
 	})
 	if err != nil {
 		return nil, err

internal/modules/homophones/module.go

@@ -72,6 +72,20 @@ func (m *Module) Propose(ctx context.Context, req contracts.ProposalRequest) ([]
 		LLMClient:      req.LLMClient,
 		Scheduler:      req.LLMScheduler,
 		DiagnosticsDir: req.DiagnosticsDir,
+		OnDroppedCandidate: func(dropped proposal_generation.DroppedCandidate) {
+			if req.OnDroppedCandidate == nil {
+				return
+			}
+			req.OnDroppedCandidate(
+				dropped.ProposalIndex,
+				dropped.TargetSegmentID,
+				dropped.OriginalText,
+				dropped.CorrectedText,
+				dropped.Confidence,
+				dropped.ReasonCode,
+				dropped.Message,
+			)
+		},
 	})
 	if err != nil {
 		return nil, err

internal/modules/spoken_word/module.go

@@ -72,6 +72,20 @@ func (m *Module) Propose(ctx context.Context, req contracts.ProposalRequest) ([]
 		LLMClient:      req.LLMClient,
 		Scheduler:      req.LLMScheduler,
 		DiagnosticsDir: req.DiagnosticsDir,
+		OnDroppedCandidate: func(dropped proposal_generation.DroppedCandidate) {
+			if req.OnDroppedCandidate == nil {
+				return
+			}
+			req.OnDroppedCandidate(
+				dropped.ProposalIndex,
+				dropped.TargetSegmentID,
+				dropped.OriginalText,
+				dropped.CorrectedText,
+				dropped.Confidence,
+				dropped.ReasonCode,
+				dropped.Message,
+			)
+		},
 	})
 	if err != nil {
 		return nil, err