fix: only use finalization prompt when plan was actually executed

Corrected finalization logic:

❌ Previous (wrong):
- Used finalization prompt for direct queries (no execution)
- Normal query path had no finalization

✅ Now (correct):
- Direct queries (no execution) → normal AI response
- Tool execution results → finalization with ownPlan:true

Logic flow:
1. If no tools executed → query_with_debug() for normal response
2. If tools executed → query_osvm_ai_with_options() with:
   - Finalization system prompt
   - ownPlan: true (prevents server from executing)
   - Formatted execution results to finalize

This ensures finalization ONLY happens after plan execution,
not for regular conversational queries.

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

Co-Authored-By: Claude <[email protected]>

Author: 0xrinegade

src/utils/streaming_agent.rs

@@ -868,17 +868,31 @@ pub async fn execute_streaming_agent(query: &str, verbose: u8) -> Result<()> {
             .collect();
 
         if tool_results_for_ai.is_empty() && tool_plan.osvm_tools_to_use.is_empty() {
-            // No tools - direct response
+            // No tools - direct response (no execution happened)
+            match ai_service.query_with_debug(query, verbose > 1).await {
+                Ok(resp) => {
+                    println!(" ✅");
+                    resp
+                }
+                Err(e) => {
+                    println!(" ❌");
+                    eprintln!("❌ Failed to generate response: {}", e);
+                    format!("Error: {}", e)
+                }
+            }
+        } else {
+            // Format results from tool execution - use finalization prompt
             let finalization_prompt = format!(
                 "You are finalizing the answer to this user query: \"{}\"\n\n\
-                 The execution has completed. Your task is to:\n\
-                 1. Format the results in a clear, human-readable way\n\
+                 Plan execution has completed. Your task is to:\n\
+                 1. Format the execution results in a clear, human-readable way\n\
                  2. Directly answer the user's question based on the data\n\
                  3. DO NOT generate any new plans or code\n\
                  4. DO NOT suggest additional steps or actions\n\
                  5. Just present the final answer clearly and concisely\n\n\
-                 Results to format:\n{:?}",
-                query, tool_results_for_ai
+                 Execution results to format:\n{}",
+                query,
+                format_plan_execution_results(&tool_plan, &tool_results_for_ai)
             );
 
             match ai_service.query_osvm_ai_with_options(
@@ -893,14 +907,11 @@ pub async fn execute_streaming_agent(query: &str, verbose: u8) -> Result<()> {
                 }
                 Err(e) => {
                     println!(" ❌");
-                    eprintln!("❌ Failed to generate response: {}", e);
-                    format!("Error: {}", e)
+                    eprintln!("❌ Failed to finalize results: {}", e);
+                    // Fallback to raw formatted results if finalization fails
+                    format_plan_execution_results(&tool_plan, &tool_results_for_ai)
                 }
             }
-        } else {
-            // Format results from tool execution
-            println!(" ✅");
-            format_plan_execution_results(&tool_plan, &tool_results_for_ai)
         }
     };