feat: add error mode for execution debugging with AI suggestions (#505)

* feat: add error mode for execution debugging with AI suggestions

Add a new `mode='error'` option to n8n_executions action=get that's optimized
for AI agents debugging workflow failures. This mode provides intelligent
error extraction with 80-99% token savings compared to `mode='full'`.

Key features:
- Error Analysis: Extracts error message, type, node name, and parameters
- Upstream Context: Samples input data from upstream node (configurable limit)
- Execution Path: Shows node execution sequence from trigger to error
- AI Suggestions: Pattern-based fix suggestions for common errors
- Workflow Fetch: Optionally fetches workflow for accurate upstream detection

New parameters for mode='error':
- errorItemsLimit (default: 2) - Sample items from upstream node
- includeStackTrace (default: false) - Full vs truncated stack trace
- includeExecutionPath (default: true) - Include node execution path
- fetchWorkflow (default: true) - Fetch workflow for upstream detection

Token efficiency:
- 11 items: ~11KB full vs ~3KB error (73% savings)
- 1001 items: ~354KB full vs ~3KB error (99% savings)

Conceived by Romuald Członkowski - https://www.aiadvisors.pl/en

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix: add security hardening to error-execution-processor

- Add prototype pollution protection (block __proto__, constructor, prototype)
- Expand sensitive data patterns (20+ patterns including JWT, OAuth, certificates)
- Create recursive sanitizeData function for deep object sanitization
- Apply sanitization to both nodeParameters and upstream sampleItems
- Add comprehensive unit tests (42 tests, 96% coverage)

Security improvements address code review findings:
- Critical: Prototype pollution protection
- Warning: Expanded sensitive data filtering
- Warning: Nested data sanitization

Concieved by Romuald Członkowski - www.aiadvisors.pl/en

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Romuald Członkowski <romualdczlonkowski@MacBook-Pro-Romuald.local>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>

src/mcp/handlers-n8n-manager.ts

@@ -1421,17 +1421,33 @@ export async function handleGetExecution(args: unknown, context?: InstanceContex
     // Parse and validate input with new parameters
     const schema = z.object({
       id: z.string(),
-      // New filtering parameters
-      mode: z.enum(['preview', 'summary', 'filtered', 'full']).optional(),
+      // Filtering parameters
+      mode: z.enum(['preview', 'summary', 'filtered', 'full', 'error']).optional(),
       nodeNames: z.array(z.string()).optional(),
       itemsLimit: z.number().optional(),
       includeInputData: z.boolean().optional(),
       // Legacy parameter (backward compatibility)
-      includeData: z.boolean().optional()
+      includeData: z.boolean().optional(),
+      // Error mode specific parameters
+      errorItemsLimit: z.number().min(0).max(100).optional(),
+      includeStackTrace: z.boolean().optional(),
+      includeExecutionPath: z.boolean().optional(),
+      fetchWorkflow: z.boolean().optional()
     });
 
     const params = schema.parse(args);
-    const { id, mode, nodeNames, itemsLimit, includeInputData, includeData } = params;
+    const {
+      id,
+      mode,
+      nodeNames,
+      itemsLimit,
+      includeInputData,
+      includeData,
+      errorItemsLimit,
+      includeStackTrace,
+      includeExecutionPath,
+      fetchWorkflow
+    } = params;
 
     /**
      * Map legacy includeData parameter to mode for backward compatibility
@@ -1470,15 +1486,33 @@ export async function handleGetExecution(args: unknown, context?: InstanceContex
       };
     }
 
+    // For error mode, optionally fetch workflow for accurate upstream detection
+    let workflow: Workflow | undefined;
+    if (effectiveMode === 'error' && fetchWorkflow !== false && execution.workflowId) {
+      try {
+        workflow = await client.getWorkflow(execution.workflowId);
+      } catch (e) {
+        // Workflow fetch failed - continue without it (use heuristics)
+        logger.debug('Could not fetch workflow for error analysis', {
+          workflowId: execution.workflowId,
+          error: e instanceof Error ? e.message : 'Unknown error'
+        });
+      }
+    }
+
     // Apply filtering using ExecutionProcessor
     const filterOptions: ExecutionFilterOptions = {
       mode: effectiveMode,
       nodeNames,
       itemsLimit,
-      includeInputData
+      includeInputData,
+      // Error mode specific options
+      errorItemsLimit,
+      includeStackTrace,
+      includeExecutionPath
     };
 
-    const processedExecution = processExecution(execution, filterOptions);
+    const processedExecution = processExecution(execution, filterOptions, workflow);
 
     return {
       success: true,