fix: stop lying to users when analysis parsing fails

When the analyzer LLM returns truncated JSON (which happens — models
hit output token limits), parseAnalysisResponse was catching the error,
printing to stderr where nobody reads it, and returning a zeroed-out
result. Every caller then printed "Analysis complete" with a green
checkmark. Your analysis didn't complete. It crashed. Tell the user.

Added a parseFailed flag to AnalysisResult so callers can distinguish
"the model scored you low" from "we couldn't even read the response."
All three call sites now show a warning with retry instructions instead
of pretending everything is fine.

Closes #56

Author: Treelovah

src/commands/analyze.ts

@@ -153,7 +153,12 @@ export const analyzeCommand = new Command('analyze')
         });
 
         saveAnalysisResult(id, analysis, getResultsDir());
-        spinner.succeed(`Analysis complete for ${id}`);
+
+        if (analysis.parseFailed) {
+          spinner.warn(`Analysis failed for ${id} — could not parse LLM response (truncated or malformed JSON)`);
+        } else {
+          spinner.succeed(`Analysis complete for ${id}`);
+        }
 
         // Print summary for single runs
         if (runIds.length === 1) {

src/commands/run.ts

@@ -349,7 +349,12 @@ export const runCommand = new Command('run')
           });
 
           const { jsonPath: analysisPath } = saveAnalysisResult(result.id, analysis, getResultsDir());
-          spinnerAnalysis.succeed('Analysis complete');
+
+          if (analysis.parseFailed) {
+            spinnerAnalysis.warn('Analysis failed — could not parse LLM response (truncated or malformed JSON). Retry with: oasis analyze ' + result.id);
+          } else {
+            spinnerAnalysis.succeed('Analysis complete');
+          }
 
           // Print analysis summary
           printAnalysisSummary(analysis);

src/interactive/run-flow.ts

@@ -552,7 +552,12 @@ export async function runBenchmarkFlow(): Promise<void> {
 
           runAnalysisResult = analysis;
           const { jsonPath: analysisPath } = saveAnalysisResult(result.id, analysis, getResultsDir());
-          spinnerAnalysis.succeed('Analysis complete');
+
+          if (analysis.parseFailed) {
+            spinnerAnalysis.warn('Analysis failed — could not parse LLM response (truncated or malformed JSON). Retry with: oasis analyze ' + result.id);
+          } else {
+            spinnerAnalysis.succeed('Analysis complete');
+          }
 
           printAnalysisSummary(analysis);
 

src/lib/analyzer.ts

@@ -334,12 +334,11 @@ async function parseAnalysisResponse(
 
     return analysisResult;
   } catch (error) {
-    console.error('Failed to parse analysis response:', error);
-
     return {
       runId,
       analyzedAt: new Date(),
       analyzerModel: DEFAULT_ANALYZER_MODEL,
+      parseFailed: true,
       attackChain: { phases: [], techniques: [], killChainCoverage: [] },
       narrative: { summary: 'Analysis parsing failed', detailed: `Error: ${error}`, keyFindings: [] },
       behavior: { approach: 'exploratory', approachDescription: 'Unable to determine', strengths: [], inefficiencies: [], decisionQuality: 0 },

src/lib/types.ts

@@ -142,6 +142,7 @@ export interface AnalysisResult {
     scoreBreakdown: string;
   };
   rubricScore?: RubricScore;
+  parseFailed?: boolean;
 }
 
 // =============================================================================