Update API URL in proxy configuration test and clean up attachment handling logic

- Modify the expected API URL in the `TestBuildProxyConfig` to include the `/v1` prefix for consistency with the updated URL building logic.
- Remove unnecessary `modified` flags in the `prepareAttachments` method to streamline attachment processing and ensure clarity in the codebase.

This change enhances the accuracy of tests and improves the maintainability of the attachment handling logic.

agent/sandbox/claude/command.go

@@ -53,8 +53,8 @@ When working with GitHub and a token is provided:
 // Only keys listed here are passed through; everything else is ignored.
 var claudeArgWhitelist = map[string]string{
 	"max_turns":        "--max-turns",        // Maximum conversation turns
-	"disallowed_tools": "--disallowed-tools",  // Comma-separated tool blacklist (e.g. "WebSearch,WebFetch")
-	"allowed_tools":    "--allowedTools",      // Comma-separated tool whitelist (e.g. "Bash,Read,Write")
+	"disallowed_tools": "--disallowed-tools", // Comma-separated tool blacklist (e.g. "WebSearch,WebFetch")
+	"allowed_tools":    "--allowedTools",     // Comma-separated tool whitelist (e.g. "Bash,Read,Write")
 }
 
 // BuildCommand builds the Claude CLI command and environment variables

agent/sandbox/claude/command_test.go

@@ -118,8 +118,9 @@ func TestBuildProxyConfig(t *testing.T) {
 
 	configStr := string(configJSON)
 	// Proxy config uses simple format
+	// BuildAPIURL adds /v1 prefix for hosts that don't end with "/"
 	assert.Contains(t, configStr, "backend")
-	assert.Contains(t, configStr, "https://api.example.com/chat/completions")
+	assert.Contains(t, configStr, "https://api.example.com/v1/chat/completions")
 	assert.Contains(t, configStr, "api_key")
 	assert.Contains(t, configStr, "key123")
 	assert.Contains(t, configStr, "model")

agent/sandbox/claude/executor.go

@@ -474,7 +474,6 @@ func (e *Executor) prepareAttachments(ctx context.Context, messages []agentConte
 
 		// Process each content part
 		var textParts []string
-		modified := false
 
 		for _, item := range parts {
 			m, ok := item.(map[string]interface{})
@@ -504,7 +503,6 @@ func (e *Executor) prepareAttachments(ctx context.Context, messages []agentConte
 				if !isWrapper {
 					// Not an attachment URL, keep as text reference
 					textParts = append(textParts, fmt.Sprintf("[Image: %s]", url))
-					modified = true
 					continue
 				}
 
@@ -513,13 +511,11 @@ func (e *Executor) prepareAttachments(ctx context.Context, messages []agentConte
 				if err != nil {
 					log.Printf("[sandbox] Warning: failed to resolve image attachment %s: %v", fileID, err)
 					textParts = append(textParts, "[Attached image: failed to load]")
-					modified = true
 					continue
 				}
 
 				textParts = append(textParts, ref)
 				hasAttachments = true
-				modified = true
 
 			case "file":
 				fileData, _ := m["file"].(map[string]interface{})
@@ -535,7 +531,6 @@ func (e *Executor) prepareAttachments(ctx context.Context, messages []agentConte
 				uploaderName, fileID, isWrapper := attachment.Parse(url)
 				if !isWrapper {
 					textParts = append(textParts, fmt.Sprintf("[File: %s]", url))
-					modified = true
 					continue
 				}
 
@@ -543,13 +538,11 @@ func (e *Executor) prepareAttachments(ctx context.Context, messages []agentConte
 				if err != nil {
 					log.Printf("[sandbox] Warning: failed to resolve file attachment %s: %v", fileID, err)
 					textParts = append(textParts, "[Attached file: failed to load]")
-					modified = true
 					continue
 				}
 
 				textParts = append(textParts, ref)
 				hasAttachments = true
-				modified = true
 
 			default:
 				// Keep other types as-is (shouldn't happen normally)
@@ -557,7 +550,11 @@ func (e *Executor) prepareAttachments(ctx context.Context, messages []agentConte
 			}
 		}
 
-		if modified && len(textParts) > 0 {
+		// Merge text parts into a single string when the original content was
+		// a multimodal array ([]interface{} / []ContentPart).  This is needed
+		// even when only "text" parts are present so that downstream code
+		// (BuildInputJSONL, etc.) always sees a plain string.
+		if len(textParts) > 0 {
 			newMsg := result[i]
 			newMsg.Content = strings.Join(textParts, "\n\n")
 			result[i] = newMsg