Improve prompt for parallel tool calls

codex-rs/core/gpt_5_codex_prompt.md

@@ -6,6 +6,12 @@ You are Codex, based on GPT-5. You are running as a coding agent in the Codex CL
 - Always set the `workdir` param when using the shell function. Do not use `cd` unless absolutely necessary.
 - When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
 
+## Exploration
+- Use `read_file` to get file contents.
+- You may output multiple `read_file` calls at the same time. Do it when you can to win in efficiency and time. This is very important.
+- Do not try to design your own parallelization strategy. Codex harness will automatically run multiple tool calls in parallel when you output them together.
+- You can only return multiple tool calls for non-mutating tools (i.e. `read-file`). Prefer doing it when possible instead of calling them once at a time to be more efficient.
+
 ## Editing constraints
 
 - Default to ASCII when editing or creating files. Only introduce non-ASCII or other Unicode characters when there is a clear justification and the file already uses them.

codex-rs/core/src/tools/parallel.rs

@@ -76,6 +76,16 @@ impl ToolCallRuntime {
         &mut self,
         output: &mut [ProcessedResponseItem],
     ) -> Result<(), CodexErr> {
+        if self.pending_calls.len() > 0 {
+            tracing::info!(
+                "Resolving {} tool calls: {:?}",
+                self.pending_calls.len(),
+                self.pending_calls
+                    .iter()
+                    .map(|call| call.index)
+                    .collect::<Vec<_>>()
+            );
+        }
         while let Some(PendingToolCall { index, handle }) = self.pending_calls.pop() {
             match handle.await {
                 Ok(Ok(response)) => {