evaluate: dedup agent prompt when comparing same agent type

When both actions are from the same agent, show the instructions once
and just compare the two report outputs + affected nodes. Saves tokens
and makes the comparison cleaner.

Co-Authored-By: Kent Overstreet <kent.overstreet@linux.dev>

poc-memory/src/cli/agent.rs

@@ -263,14 +263,32 @@ fn llm_compare(
     b: &(String, String, String),
     model: &str,
 ) -> Result<std::cmp::Ordering, String> {
-    let prompt = format!(
+    let prompt = if a.0 == b.0 {
+        // Same agent type — show instructions once
+        // Split context at "## Report output" to extract shared prompt
+        let split_a: Vec<&str> = a.2.splitn(2, "## Report output").collect();
+        let split_b: Vec<&str> = b.2.splitn(2, "## Report output").collect();
+        let shared_prompt = split_a.first().unwrap_or(&"");
+        let report_a = split_a.get(1).unwrap_or(&"");
+        let report_b = split_b.get(1).unwrap_or(&"");
+        format!(
+            "Compare two actions from the same {} agent. Which was better?\n\n\
+             {}\n\n\
+             ## Action A\n## Report output{}\n\n\
+             ## Action B\n## Report output{}\n\n\
+             Reply with ONLY: BETTER: A  or  BETTER: B  or  BETTER: TIE",
+            a.0, shared_prompt, report_a, report_b
+        )
+    } else {
+        format!(
             "Compare these two memory graph agent actions. Which one was better \
              for building a useful, well-organized knowledge graph?\n\n\
              ## Action A ({} agent)\n{}\n\n\
              ## Action B ({} agent)\n{}\n\n\
              Reply with ONLY: BETTER: A  or  BETTER: B  or  BETTER: TIE",
             a.0, a.2, b.0, b.2
-    );
+        )
+    };
 
     let response = if model == "haiku" {
         llm::call_haiku("compare", &prompt)?