Fix: Restore visibility of tool execution output

README.md

@@ -35,6 +35,9 @@ mycoder "Implement a React component that displays a list of items"
 # Run with a prompt from a file
 mycoder -f prompt.txt
 
+# Enable interactive corrections during execution (press Ctrl+M to send corrections)
+mycoder --interactive "Implement a React component that displays a list of items"
+
 # Disable user prompts for fully automated sessions
 mycoder --userPrompt false "Generate a basic Express.js server"
 
@@ -98,7 +101,7 @@ export default {
 
   // Base URL configuration (for providers that need it)
   baseUrl: 'http://localhost:11434', // Example for Ollama
-  
+
   // MCP configuration
   mcp: {
     servers: [
@@ -119,6 +122,35 @@ export default {
 
 CLI arguments will override settings in your configuration file.
 
+## Interactive Corrections
+
+MyCoder supports sending corrections to the main agent while it's running. This is useful when you notice the agent is going off track or needs additional information.
+
+### Usage
+
+1. Start MyCoder with the `--interactive` flag:
+   ```bash
+   mycoder --interactive "Implement a React component"
+   ```
+
+2. While the agent is running, press `Ctrl+M` to enter correction mode
+3. Type your correction or additional context
+4. Press Enter to send the correction to the agent
+
+The agent will receive your message and incorporate it into its decision-making process, similar to how parent agents can send messages to sub-agents.
+
+### Configuration
+
+You can enable interactive corrections in your configuration file:
+
+```js
+// mycoder.config.js
+export default {
+  // ... other options
+  interactive: true,
+};
+```
+
 ### GitHub Comment Commands
 
 MyCoder can be triggered directly from GitHub issue comments using the flexible `/mycoder` command:
@@ -182,4 +214,4 @@ Please see [CONTRIBUTING.md](CONTRIBUTING.md) for details on how to contribute t
 
 ## License
 
-This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
+This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

packages/agent/README.md

@@ -69,29 +69,36 @@ MyCoder Agent supports the Model Context Protocol:
 ## Available Tools
 
 ### File & Text Manipulation
+
 - **textEditor**: View, create, and edit files with persistent state
   - Commands: view, create, str_replace, insert, undo_edit
   - Line number support and partial file viewing
 
 ### System Interaction
+
 - **shellStart**: Execute shell commands with sync/async modes
 - **shellMessage**: Interact with running shell processes
 - **shellExecute**: One-shot shell command execution
 - **listShells**: List all running shell processes
 
 ### Agent Management
+
 - **agentStart**: Create sub-agents for parallel tasks
-- **agentMessage**: Send messages to sub-agents
+- **agentMessage**: Send messages to sub-agents and retrieve their output (including captured logs)
 - **agentDone**: Complete the current agent's execution
 - **listAgents**: List all running agents
 
+The agent system automatically captures log, warn, and error messages from agents and their immediate tools, which are included in the output returned by agentMessage.
+
 ### Network & Web
+
 - **fetch**: Make HTTP requests to APIs
 - **sessionStart**: Start browser automation sessions
 - **sessionMessage**: Control browser sessions (navigation, clicking, typing)
 - **listSessions**: List all browser sessions
 
 ### Utility Tools
+
 - **sleep**: Pause execution for a specified duration
 - **userPrompt**: Request input from the user
 
@@ -145,10 +152,10 @@ const tools = [textEditorTool, shellStartTool];
 
 // Run the agent
 const result = await toolAgent(
-  "Write a simple Node.js HTTP server and save it to server.js",
+  'Write a simple Node.js HTTP server and save it to server.js',
   tools,
   {
-    getSystemPrompt: () => "You are a helpful coding assistant...",
+    getSystemPrompt: () => 'You are a helpful coding assistant...',
     maxIterations: 10,
   },
   {
@@ -157,7 +164,7 @@ const result = await toolAgent(
     model: 'claude-3-opus-20240229',
     apiKey: process.env.ANTHROPIC_API_KEY,
     workingDirectory: process.cwd(),
-  }
+  },
 );
 
 console.log('Agent result:', result);
@@ -169,4 +176,4 @@ We welcome contributions! Please see our [CONTRIBUTING.md](../CONTRIBUTING.md) f
 
 ## License
 
-MIT
+MIT

packages/agent/src/core/toolAgent/toolAgentCore.test.ts

@@ -7,7 +7,7 @@ describe('toolAgentCore empty response detection', () => {
     const fileContent = `
     if (!text.length && toolCalls.length === 0) {
       // Only consider it empty if there's no text AND no tool calls
-      logger.verbose('Received truly empty response from agent (no text and no tool calls), sending reminder');
+      logger.debug('Received truly empty response from agent (no text and no tool calls), sending reminder');
       messages.push({
         role: 'user',
         content: [