Add agent run inspection debugging tip

Author: 张安哲

README.md

@@ -87,6 +87,10 @@ Vibe coding is a mindset for AI assisted development where you "vibe" with the A
     - If the first fix doesn't work, explain the outcome and try again. 
     - The AI is often capable of resolving it.
 
+- **Inspect Agent Runs**:
+    - When an AI task is slow, costly, or repeatedly failing, review the saved session logs instead of judging only the final answer.
+    - Tools like [agenttrace](https://github.com/luoyuctl/agenttrace) can surface token usage, cost, latency gaps, tool failures, and retry loops across Claude Code, Codex CLI, Gemini CLI, Cursor exports, and similar agents.
+
 - **Iterate and Refine**:
     - Embrace rapid iteration.
     - Don't worry about perfect designs initially.
@@ -174,4 +178,4 @@ This project is licensed under [MIT License](LICENSE).
 
 ---
 
-⭐️ If you find this repository helpful, please consider giving it a star!
+⭐️ If you find this repository helpful, please consider giving it a star!

testing_and_debugging_guide/README.md

@@ -9,7 +9,11 @@
     - If the first fix doesn't work, explain the outcome and try again. 
     - The AI is often capable of resolving it.
 
+- **Inspect Agent Runs**:
+    - When an AI task is slow, costly, or repeatedly failing, review the saved session logs instead of judging only the final answer.
+    - Tools like [agenttrace](https://github.com/luoyuctl/agenttrace) can surface token usage, cost, latency gaps, tool failures, and retry loops across Claude Code, Codex CLI, Gemini CLI, Cursor exports, and similar agents.
+
 - **Iterate and Refine**:
     - Embrace rapid iteration.
     - Don't worry about perfect designs initially.
-    - Improve them step by step.
+    - Improve them step by step.