diff --git a/internal/cli/run.go b/internal/cli/run.go
index a38cfa3..742655e 100644
--- a/internal/cli/run.go
+++ b/internal/cli/run.go
@@ -9,6 +9,7 @@ import (
 	"time"
 
 	"gitea.maximumdirect.net/eric/audita/internal/core/config"
+	"gitea.maximumdirect.net/eric/audita/internal/core/diagnostics"
 	"gitea.maximumdirect.net/eric/audita/internal/core/io"
 	"gitea.maximumdirect.net/eric/audita/internal/core/normalization"
 	"gitea.maximumdirect.net/eric/audita/internal/core/reporting"
@@ -24,25 +25,49 @@ type processInvocation struct {
 }
 
 var processRunner = func(inv processInvocation, stdout io.Writer) (*normalization.NormalizationSummary, error) {
+	// Create run directory early to capture artifacts
+	runDir, err := diagnostics.NewRunDirectory(inv.Config.WorkDir, string(inv.Config.WorkDirRetention))
+	if err != nil {
+		return nil, fmt.Errorf("run_dir_creation: %w", err)
+	}
+	defer func() {
+		// Apply retention policy - failed runs are always kept
+		// For now, we consider any run that saved source as successful
+		if runDir != nil {
+			_ = runDir.ApplyRetention()
+		}
+	}()
+
 	transcriptBytes, err := io.ReadRequiredFile(inv.TranscriptPath, "transcript")
 	if err != nil {
+		// Best effort to write error log on failure
+		_ = runDir.WriteErrorLog(fmt.Sprintf("transcript_read: %v", err))
 		return nil, fmt.Errorf("transcript_read: %w", err)
 	}
 
 	glossaryBytes, err := io.ReadRequiredFile(inv.GlossaryPath, "glossary")
 	if err != nil {
+		_ = runDir.WriteErrorLog(fmt.Sprintf("glossary_read: %v", err))
 		return nil, fmt.Errorf("glossary_read: %w", err)
 	}
 
 	// Parse and validate transcript using typed schema
 	transcript, err := schema.ParseSourceTranscriptJSON(transcriptBytes)
 	if err != nil {
+		_ = runDir.WriteErrorLog(fmt.Sprintf("transcript_schema: %v", err))
 		return nil, fmt.Errorf("transcript_schema: %w", err)
 	}
 
+	// Write source transcript artifacts
+	if err := runDir.WriteSourceTranscript(transcript, transcriptBytes); err != nil {
+		_ = runDir.WriteErrorLog(fmt.Sprintf("source_artifact: %v", err))
+		// Continue without failing the whole process
+	}
+
 	// Parse and validate glossary using typed schema
 	glossary, err := schema.ParseGlossaryYAML(glossaryBytes)
 	if err != nil {
+		_ = runDir.WriteErrorLog(fmt.Sprintf("glossary_schema: %v", err))
 		return nil, fmt.Errorf("glossary_schema: %w", err)
 	}
 
@@ -75,21 +100,36 @@ var processRunner = func(inv processInvocation, stdout io.Writer) (*normalizatio
 
 	normalizedTranscript, normalizationSummary := normalizer.Normalize(normalizedTranscript)
 
+	// Write normalized transcript artifact
+	if err := runDir.WriteNormalizedTranscript(normalizedTranscript); err != nil {
+		_ = runDir.WriteErrorLog(fmt.Sprintf("normalized_artifact: %v", err))
+		// Continue without failing the whole process
+	}
+
+	// Write normalization summary artifact
+	if err := runDir.WriteNormalizationSummary(normalizationSummary); err != nil {
+		_ = runDir.WriteErrorLog(fmt.Sprintf("normalization_summary: %v", err))
+		// Continue without failing the whole process
+	}
+
 	// Serialize normalized transcript to JSON
 	outputBytes, err := schema.TranscriptToJSON(normalizedTranscript)
 	if err != nil {
-		return nil, fmt.Errorf("failed to serialize transcript: %w", err)
+		_ = runDir.WriteErrorLog(fmt.Sprintf("serialization: %v", err))
+		return nil, fmt.Errorf("serialization: %w", err)
 	}
 
 	if strings.TrimSpace(inv.OutputPath) != "" {
 		if err := io.WriteFile(inv.OutputPath, outputBytes); err != nil {
+			_ = runDir.WriteErrorLog(fmt.Sprintf("output_write: %v", err))
 			return nil, err
 		}
 		return normalizationSummary, nil
 	}
 
 	if _, err := stdout.Write(outputBytes); err != nil {
-		return nil, fmt.Errorf("failed to write transcript to stdout: %w", err)
+		_ = runDir.WriteErrorLog(fmt.Sprintf("stdout_write: %v", err))
+		return nil, fmt.Errorf("stdout_write: %w", err)
 	}
 	return normalizationSummary, nil
 }
diff --git a/internal/cli/run_test.go b/internal/cli/run_test.go
index 91ed2ba..e995e9a 100644
--- a/internal/cli/run_test.go
+++ b/internal/cli/run_test.go
@@ -808,48 +808,327 @@ func TestRunProcessReportFailedGlossarySchema(t *testing.T) {
 	}
 }
 
-func TestRunProcessReportJSONNotInStdout(t *testing.T) {
+func TestRunProcessRunDirectoryCreation(t *testing.T) {
 	var stdout bytes.Buffer
 	var stderr bytes.Buffer
 
-	transcriptJSON := `[
-		{"id": 1, "speaker": "Alice", "start": 0.0, "end": 1.0, "text": "Hello"}
-	]`
+	transcriptJSON := `[{"id": 1, "speaker": "Alice", "start": 0.0, "end": 1.0, "text": "Hello"}]`
 
 	transcriptPath := filepath.Join(t.TempDir(), "transcript.json")
 	glossaryPath := fixturePath("tiny_glossary.yaml")
-	reportPath := filepath.Join(t.TempDir(), "report.json")
+	workDir := filepath.Join(t.TempDir(), "work")
 
 	if err := os.WriteFile(transcriptPath, []byte(transcriptJSON), 0644); err != nil {
 		t.Fatalf("failed to create test transcript: %v", err)
 	}
 
-	exitCode := Run([]string{"process", transcriptPath, "--glossary", glossaryPath, "--report-json", reportPath}, &stdout, &stderr)
+	exitCode := Run([]string{"process", transcriptPath, "--glossary", glossaryPath, "--work-dir", workDir}, &stdout, &stderr)
 	if exitCode != 0 {
 		t.Fatalf("expected exit code 0, got %d with stderr %q", exitCode, stderr.String())
 	}
 
-	// Read the report to get its content
-	reportBytes := readFile(t, reportPath)
-
-	// Verify report JSON is not in stdout
-	if bytes.Contains(stdout.Bytes(), reportBytes) {
-		t.Fatalf("stdout was polluted with report JSON")
+	// Check that run directory was created
+	runDirs, err := os.ReadDir(workDir)
+	if err != nil {
+		t.Fatalf("failed to read work directory: %v", err)
+	}
+	if len(runDirs) != 1 {
+		t.Fatalf("expected 1 run directory, got %d", len(runDirs))
 	}
 
-	// Verify stdout only contains transcript JSON
-	if !json.Valid(stdout.Bytes()) {
-		t.Fatalf("expected valid JSON in stdout, got %q", stdout.String())
+	runPath := filepath.Join(workDir, runDirs[0].Name())
+	
+	// Check that source transcript artifact exists
+	if _, err := os.Stat(filepath.Join(runPath, "source-transcript.json")); err != nil {
+		t.Fatalf("expected source transcript artifact: %v", err)
 	}
 
-	// Parse stdout to verify it's a transcript, not a report
-	var stdoutContent map[string]interface{}
-	if err := json.Unmarshal(stdout.Bytes(), &stdoutContent); err != nil {
-		t.Fatalf("failed to parse stdout as JSON: %v", err)
+	// Check that normalized transcript artifact exists
+	if _, err := os.Stat(filepath.Join(runPath, "normalized-transcript.json")); err != nil {
+		t.Fatalf("expected normalized transcript artifact: %v", err)
 	}
 
-	// Transcript should be an array at top level
-	if _, isArray := stdoutContent["segments"]; !isArray && stdoutContent["segments"] != nil {
-		t.Fatalf("expected transcript in stdout, got report structure")
+	// Check that normalization summary artifact exists
+	if _, err := os.Stat(filepath.Join(runPath, "normalization-summary.json")); err != nil {
+		t.Fatalf("expected normalization summary artifact: %v", err)
+	}
+
+	// Check that report artifact exists
+	if _, err := os.Stat(filepath.Join(runPath, "report.json")); err != nil {
+		t.Fatalf("expected report artifact: %v", err)
+	}
+}
+
+func TestRunProcessSourceTranscriptArtifact(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	transcriptJSON := `[{"id": 1, "speaker": "Alice", "start": 0.0, "end": 1.0, "text": "Hello"}]`
+
+	transcriptPath := filepath.Join(t.TempDir(), "transcript.json")
+	glossaryPath := fixturePath("tiny_glossary.yaml")
+	workDir := filepath.Join(t.TempDir(), "work")
+
+	if err := os.WriteFile(transcriptPath, []byte(transcriptJSON), 0644); err != nil {
+		t.Fatalf("failed to create test transcript: %v", err)
+	}
+
+	exitCode := Run([]string{"process", transcriptPath, "--glossary", glossaryPath, "--work-dir", workDir}, &stdout, &stderr)
+	if exitCode != 0 {
+		t.Fatalf("expected exit code 0, got %d with stderr %q", exitCode, stderr.String())
+	}
+
+	// Find the run directory
+	runDirs, err := os.ReadDir(workDir)
+	if err != nil {
+		t.Fatalf("failed to read work directory: %v", err)
+	}
+	runPath := filepath.Join(workDir, runDirs[0].Name())
+
+	// Read and verify source transcript artifact
+	sourceBytes := readFile(t, filepath.Join(runPath, "source-transcript.json"))
+	if !bytes.Equal(sourceBytes, []byte(transcriptJSON)) {
+		t.Fatalf("source transcript artifact doesn't match input")
+	}
+}
+
+func TestRunProcessNormalizedTranscriptArtifact(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	transcriptJSON := `[
+		{"id": 1, "speaker": "Alice", "start": 0.0, "end": 1.0, "text": "Hello"},
+		{"id": 2, "speaker": "Alice", "start": 1.5, "end": 2.5, "text": "world"}
+	]`
+
+	transcriptPath := filepath.Join(t.TempDir(), "transcript.json")
+	glossaryPath := fixturePath("tiny_glossary.yaml")
+	workDir := filepath.Join(t.TempDir(), "work")
+
+	if err := os.WriteFile(transcriptPath, []byte(transcriptJSON), 0644); err != nil {
+		t.Fatalf("failed to create test transcript: %v", err)
+	}
+
+	exitCode := Run([]string{"process", transcriptPath, "--glossary", glossaryPath, "--work-dir", workDir}, &stdout, &stderr)
+	if exitCode != 0 {
+		t.Fatalf("expected exit code 0, got %d with stderr %q", exitCode, stderr.String())
+	}
+
+	// Find the run directory
+	runDirs, err := os.ReadDir(workDir)
+	if err != nil {
+		t.Fatalf("failed to read work directory: %v", err)
+	}
+	runPath := filepath.Join(workDir, runDirs[0].Name())
+
+	// Read and verify normalized transcript artifact
+	normalizedBytes := readFile(t, filepath.Join(runPath, "normalized-transcript.json"))
+	var normalizedTranscript schema.Transcript
+	if err := schema.ParseTranscriptJSON(normalizedBytes); err != nil {
+		t.Fatalf("failed to parse normalized transcript artifact: %v", err)
+	}
+
+	// Should be merged (same speaker, small gap)
+	if len(normalizedTranscript.Segments) != 1 {
+		t.Fatalf("expected 1 merged segment in normalized artifact, got %d", len(normalizedTranscript.Segments))
+	}
+}
+
+func TestRunProcessNormalizationSummaryArtifact(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	transcriptJSON := `[
+		{"id": 1, "speaker": "Alice", "start": 0.0, "end": 1.0, "text": "Hello"},
+		{"id": 2, "speaker": "Alice", "start": 1.5, "end": 2.5, "text": "world"}
+	]`
+
+	transcriptPath := filepath.Join(t.TempDir(), "transcript.json")
+	glossaryPath := fixturePath("tiny_glossary.yaml")
+	workDir := filepath.Join(t.TempDir(), "work")
+
+	if err := os.WriteFile(transcriptPath, []byte(transcriptJSON), 0644); err != nil {
+		t.Fatalf("failed to create test transcript: %v", err)
+	}
+
+	exitCode := Run([]string{"process", transcriptPath, "--glossary", glossaryPath, "--work-dir", workDir}, &stdout, &stderr)
+	if exitCode != 0 {
+		t.Fatalf("expected exit code 0, got %d with stderr %q", exitCode, stderr.String())
+	}
+
+	// Find the run directory
+	runDirs, err := os.ReadDir(workDir)
+	if err != nil {
+		t.Fatalf("failed to read work directory: %v", err)
+	}
+	runPath := filepath.Join(workDir, runDirs[0].Name())
+
+	// Read and verify normalization summary artifact
+	summaryBytes := readFile(t, filepath.Join(runPath, "normalization-summary.json"))
+	var summary normalization.NormalizationSummary
+	if err := json.Unmarshal(summaryBytes, &summary); err != nil {
+		t.Fatalf("failed to parse normalization summary: %v", err)
+	}
+
+	// Verify summary content
+	if summary.InputSegmentCount != 2 {
+		t.Fatalf("expected input count 2, got %d", summary.InputSegmentCount)
+	}
+	if summary.OutputSegmentCount != 1 {
+		t.Fatalf("expected output count 1, got %d", summary.OutputSegmentCount)
+	}
+	if summary.MergesPerformed != 1 {
+		t.Fatalf("expected 1 merge, got %d", summary.MergesPerformed)
+	}
+}
+
+func TestRunProcessReportArtifact(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	transcriptJSON := `[{"id": 1, "speaker": "Alice", "start": 0.0, "end": 1.0, "text": "Hello"}]`
+
+	transcriptPath := filepath.Join(t.TempDir(), "transcript.json")
+	glossaryPath := fixturePath("tiny_glossary.yaml")
+	workDir := filepath.Join(t.TempDir(), "work")
+
+	if err := os.WriteFile(transcriptPath, []byte(transcriptJSON), 0644); err != nil {
+		t.Fatalf("failed to create test transcript: %v", err)
+	}
+
+	exitCode := Run([]string{"process", transcriptPath, "--glossary", glossaryPath, "--work-dir", workDir}, &stdout, &stderr)
+	if exitCode != 0 {
+		t.Fatalf("expected exit code 0, got %d with stderr %q", exitCode, stderr.String())
+	}
+
+	// Find the run directory
+	runDirs, err := os.ReadDir(workDir)
+	if err != nil {
+		t.Fatalf("failed to read work directory: %v", err)
+	}
+	runPath := filepath.Join(workDir, runDirs[0].Name())
+
+	// Verify report artifact exists and is valid
+	reportBytes := readFile(t, filepath.Join(runPath, "report.json"))
+	var report reporting.ProcessReport
+	if err := json.Unmarshal(reportBytes, &report); err != nil {
+		t.Fatalf("failed to parse report: %v", err)
+	}
+
+	if report.Status != "success" {
+		t.Fatalf("expected success status, got %q", report.Status)
+	}
+}
+
+func TestRunProcessErrorLogOnFailure(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	workDir := filepath.Join(t.TempDir(), "work")
+
+	exitCode := Run([]string{
+		"process",
+		fixturePath("transcript_empty_speaker.json"),
+		"--glossary",
+		fixturePath("tiny_glossary.yaml"),
+		"--work-dir",
+		workDir,
+	}, &stdout, &stderr)
+	if exitCode == 0 {
+		t.Fatalf("expected nonzero exit code for invalid transcript")
+	}
+
+	// Find the run directory
+	runDirs, err := os.ReadDir(workDir)
+	if err != nil {
+		t.Fatalf("failed to read work directory: %v", err)
+	}
+	if len(runDirs) != 1 {
+		t.Fatalf("expected 1 run directory on failure, got %d", len(runDirs))
+	}
+
+	runPath := filepath.Join(workDir, runDirs[0].Name())
+
+	// Verify error.log exists
+	if _, err := os.Stat(filepath.Join(runPath, "error.log")); err != nil {
+		t.Fatalf("expected error.log on failure: %v", err)
+	}
+
+	// Verify error.log contains useful information
+	errorBytes := readFile(t, filepath.Join(runPath, "error.log"))
+	if !strings.Contains(string(errorBytes), "transcript_schema") {
+		t.Fatalf("expected error phase in error.log, got %q", string(errorBytes))
+	}
+}
+
+func TestRunProcessFailedRunsRetained(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	workDir := filepath.Join(t.TempDir(), "work")
+
+	exitCode := Run([]string{
+		"process",
+		fixturePath("transcript_empty_speaker.json"),
+		"--glossary",
+		fixturePath("tiny_glossary.yaml"),
+		"--work-dir",
+		workDir,
+		"--work-dir-retention",
+		"never", // Even with "never", failed runs should be retained
+	}, &stdout, &stderr)
+	if exitCode == 0 {
+		t.Fatalf("expected nonzero exit code for invalid transcript")
+	}
+
+	// Find the run directory
+	runDirs, err := os.ReadDir(workDir)
+	if err != nil {
+		t.Fatalf("failed to read work directory: %v", err)
+	}
+	if len(runDirs) != 1 {
+		t.Fatalf("expected failed run to be retained, got %d run directories", len(runDirs))
+	}
+
+	// Failed runs should always be retained regardless of retention policy
+	runPath := filepath.Join(workDir, runDirs[0].Name())
+	if _, err := os.Stat(runPath); err != nil {
+		t.Fatalf("failed run directory should be retained: %v", err)
+	}
+}
+
+func TestRunProcessSuccessfulRunRetention(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	transcriptJSON := `[{"id": 1, "speaker": "Alice", "start": 0.0, "end": 1.0, "text": "Hello"}]`
+
+	transcriptPath := filepath.Join(t.TempDir(), "transcript.json")
+	glossaryPath := fixturePath("tiny_glossary.yaml")
+	workDir := filepath.Join(t.TempDir(), "work")
+
+	if err := os.WriteFile(transcriptPath, []byte(transcriptJSON), 0644); err != nil {
+		t.Fatalf("failed to create test transcript: %v", err)
+	}
+
+	// Test with "never" retention - successful runs should be removed
+	exitCode := Run([]string{
+		"process",
+		transcriptPath,
+		"--glossary",
+		glossaryPath,
+		"--work-dir",
+		workDir,
+		"--work-dir-retention",
+		"never",
+	}, &stdout, &stderr)
+	if exitCode != 0 {
+		t.Fatalf("expected exit code 0, got %d with stderr %q", exitCode, stderr.String())
+	}
+
+	// With "never" retention, successful runs should be removed
+	runDirs, err := os.ReadDir(workDir)
+	if err == nil && len(runDirs) > 0 {
+		t.Fatalf("expected successful run to be removed with 'never' retention, found %d directories", len(runDirs))
 	}
 }
diff --git a/internal/core/diagnostics/run_dir.go b/internal/core/diagnostics/run_dir.go
new file mode 100644
index 0000000..277aeb2
--- /dev/null
+++ b/internal/core/diagnostics/run_dir.go
@@ -0,0 +1,145 @@
+package diagnostics
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+	"path/filepath"
+	"time"
+
+	"gitea.maximumdirect.net/eric/audita/internal/core/normalization"
+	"gitea.maximumdirect.net/eric/audita/internal/core/reporting"
+	"gitea.maximumdirect.net/eric/audita/internal/core/schema"
+)
+
+// RunDirectory represents a per-run diagnostics directory
+type RunDirectory struct {
+	path        string
+	retention   string
+	createdAt   time.Time
+	sourceSaved bool
+}
+
+// NewRunDirectory creates a new run directory under the configured work dir
+func NewRunDirectory(workDir, retention string) (*RunDirectory, error) {
+	if workDir == "" {
+		workDir = ".audita-runs"
+	}
+
+	// Create work directory if it doesn't exist
+	if err := os.MkdirAll(workDir, 0o755); err != nil {
+		return nil, fmt.Errorf("failed to create work directory %q: %w", workDir, err)
+	}
+
+	// Create unique run directory with timestamp
+	timestamp := time.Now().UTC().Format("2006-01-02T15-04-05Z")
+	runID := fmt.Sprintf("run-%s", timestamp)
+	runPath := filepath.Join(workDir, runID)
+
+	if err := os.Mkdir(runPath, 0o755); err != nil {
+		return nil, fmt.Errorf("failed to create run directory %q: %w", runPath, err)
+	}
+
+	return &RunDirectory{
+		path:      runPath,
+		retention: retention,
+		createdAt: time.Now().UTC(),
+	}, nil
+}
+
+// Path returns the run directory path
+func (r *RunDirectory) Path() string {
+	return r.path
+}
+
+// WriteSourceTranscript writes the source transcript artifact
+func (r *RunDirectory) WriteSourceTranscript(transcript *schema.SourceTranscript, raw []byte) error {
+	// Write raw source for reference
+	sourcePath := filepath.Join(r.path, "source-transcript.json")
+	if err := os.WriteFile(sourcePath, raw, 0o644); err != nil {
+		return fmt.Errorf("failed to write source transcript: %w", err)
+	}
+
+	// Write parsed source for debugging
+	parsedPath := filepath.Join(r.path, "source-transcript-parsed.json")
+	parsedBytes, err := json.MarshalIndent(transcript, "", "  ")
+	if err != nil {
+		return fmt.Errorf("failed to marshal parsed source transcript: %w", err)
+	}
+	parsedBytes = append(parsedBytes, '\n')
+	if err := os.WriteFile(parsedPath, parsedBytes, 0o644); err != nil {
+		return fmt.Errorf("failed to write parsed source transcript: %w", err)
+	}
+
+	r.sourceSaved = true
+	return nil
+}
+
+// WriteNormalizedTranscript writes the normalized transcript artifact
+func (r *RunDirectory) WriteNormalizedTranscript(transcript *schema.Transcript) error {
+	normalizedPath := filepath.Join(r.path, "normalized-transcript.json")
+	bytes, err := schema.TranscriptToJSON(transcript)
+	if err != nil {
+		return fmt.Errorf("failed to serialize normalized transcript: %w", err)
+	}
+	if err := os.WriteFile(normalizedPath, bytes, 0o644); err != nil {
+		return fmt.Errorf("failed to write normalized transcript: %w", err)
+	}
+	return nil
+}
+
+// WriteNormalizationSummary writes the normalization summary artifact
+func (r *RunDirectory) WriteNormalizationSummary(summary *normalization.NormalizationSummary) error {
+	summaryPath := filepath.Join(r.path, "normalization-summary.json")
+	bytes, err := json.MarshalIndent(summary, "", "  ")
+	if err != nil {
+		return fmt.Errorf("failed to marshal normalization summary: %w", err)
+	}
+	bytes = append(bytes, '\n')
+	if err := os.WriteFile(summaryPath, bytes, 0o644); err != nil {
+		return fmt.Errorf("failed to write normalization summary: %w", err)
+	}
+	return nil
+}
+
+// WriteReport writes the authoritative report artifact
+func (r *RunDirectory) WriteReport(report reporting.ProcessReport) error {
+	reportPath := filepath.Join(r.path, "report.json")
+	bytes, err := json.MarshalIndent(report, "", "  ")
+	if err != nil {
+		return fmt.Errorf("failed to marshal report: %w", err)
+	}
+	bytes = append(bytes, '\n')
+	if err := os.WriteFile(reportPath, bytes, 0o644); err != nil {
+		return fmt.Errorf("failed to write report: %w", err)
+	}
+	return nil
+}
+
+// WriteErrorLog writes an error log on failure
+func (r *RunDirectory) WriteErrorLog(errorMessage string) error {
+	errorPath := filepath.Join(r.path, "error.log")
+	return os.WriteFile(errorPath, []byte(errorMessage+"\n"), 0o644)
+}
+
+// ApplyRetention applies the retention policy
+// TODO: In later phases, implement full retention logic including:
+// - auto mode should keep runs with skipped corrections
+// - Add proper cleanup of old runs
+// - Consider disk space limits
+func (r *RunDirectory) ApplyRetention() error {
+	// For Phase 2, implement minimal retention:
+	// - always: keep everything (do nothing)
+	// - never: remove successful runs
+	// - auto: remove successful runs (same as never for now)
+	// - failed runs are always kept (handled by caller)
+
+	if r.retention == "never" || r.retention == "auto" {
+		// Remove successful runs
+		if r.sourceSaved { // Simple heuristic: if we saved source, it was successful
+			return os.RemoveAll(r.path)
+		}
+	}
+	// always: keep everything, or failed runs: keep everything
+	return nil
+}