fix: use markdown block type instead of mrkdwn for agent responses

Switch response blocks from section+mrkdwn to the markdown block type
so that standard Markdown formatting from the LLM renders correctly in
Slack. Update tool return values and system prompts to use standard
Markdown syntax (e.g. **bold** instead of *bold*).

Author: mwbrooks

openai-agents-sdk/agent/support_agent.py

@@ -20,7 +20,7 @@
 - Confident but honest when you don't know something
 
 ## Formatting Rules
-- Use Slack mrkdwn syntax: *bold*, _italic_, `code`, ```code blocks```, > blockquotes
+- Use standard Markdown syntax: **bold**, _italic_, `code`, ```code blocks```, > blockquotes
 - Use bullet points for multi-step instructions
 - Keep responses concise — aim for helpful, not verbose
 - When referencing ticket IDs or system names, use `inline code`

openai-agents-sdk/agent/tools/knowledge_base.py

@@ -122,7 +122,7 @@ async def search_knowledge_base(
     results = []
     for article in matches:
         results.append(
-            f"*{article['title']}* ({article['article_id']})\n{article['content']}"
+            f"**{article['title']}** ({article['article_id']})\n{article['content']}"
         )
 
     return "\n\n---\n\n".join(results)

openai-agents-sdk/agent/tools/password_reset.py

@@ -18,7 +18,7 @@ async def trigger_password_reset(
         target_user: The username or email of the user whose password should be reset.
     """
     return (
-        f"Password reset initiated for *{target_user}*.\n\n"
+        f"Password reset initiated for **{target_user}**.\n\n"
         f"A reset link has been sent to the email address on file. "
         f"The link will expire in 30 minutes.\n\n"
         f"_If the user doesn't receive the email within 5 minutes, "

openai-agents-sdk/agent/tools/system_status.py

@@ -76,7 +76,7 @@ async def check_system_status(
             }.get(info["status"], ":white_circle:")
 
             return (
-                f"*{info['name']}* {status_emoji} `{info['status'].upper()}`\n"
+                f"**{info['name']}** {status_emoji} `{info['status'].upper()}`\n"
                 f"{info['details']}"
             )
 

openai-agents-sdk/agent/tools/ticket.py

@@ -29,11 +29,11 @@ async def create_support_ticket(
 
     return (
         f"Support ticket created successfully.\n"
-        f"*Ticket ID:* {ticket_id}\n"
-        f"*Title:* {title}\n"
-        f"*Priority:* {priority}\n"
-        f"*Category:* {category}\n"
-        f"*Status:* Open\n"
-        f"*Assigned to:* IT Support Queue\n\n"
+        f"**Ticket ID:** {ticket_id}\n"
+        f"**Title:** {title}\n"
+        f"**Priority:** {priority}\n"
+        f"**Category:** {category}\n"
+        f"**Status:** Open\n"
+        f"**Assigned to:** IT Support Queue\n\n"
         f"The IT team will review this ticket and follow up within the SLA for {priority} priority issues."
     )

openai-agents-sdk/agent/tools/user_permissions.py

@@ -18,12 +18,12 @@ async def lookup_user_permissions(
         system: The system or resource to check permissions for (e.g., 'github', 'jira', 'aws').
     """
     return (
-        f"*Permissions for {target_user} on {system}:*\n\n"
-        f"*Groups:* `{system}-users`, `{system}-readonly`\n"
-        f"*Access Level:* Standard User\n"
-        f"*Last Login:* 2 hours ago\n"
-        f"*Account Status:* Active\n"
-        f"*MFA Enabled:* Yes\n\n"
+        f"**Permissions for {target_user} on {system}:**\n\n"
+        f"**Groups:** `{system}-users`, `{system}-readonly`\n"
+        f"**Access Level:** Standard User\n"
+        f"**Last Login:** 2 hours ago\n"
+        f"**Account Status:** Active\n"
+        f"**MFA Enabled:** Yes\n\n"
         f"_To request elevated access, the user's manager must submit an access request "
         f"through the IT portal._"
     )

openai-agents-sdk/listeners/events/app_mentioned.py

@@ -83,11 +83,8 @@ def handle_app_mentioned(client: WebClient, event: dict, logger: Logger, say: Sa
         feedback_blocks = create_feedback_block()
         response_blocks = [
             {
-                "type": "section",
-                "text": {
-                    "type": "mrkdwn",
-                    "text": result.final_output,
-                },
+                "type": "markdown",
+                "text": result.final_output,
             },
             *feedback_blocks,
         ]

openai-agents-sdk/listeners/events/message_im.py

@@ -81,11 +81,8 @@ def handle_message_im(client: WebClient, event: dict, logger: Logger, say: Say):
         feedback_blocks = create_feedback_block()
         response_blocks = [
             {
-                "type": "section",
-                "text": {
-                    "type": "mrkdwn",
-                    "text": result.final_output,
-                },
+                "type": "markdown",
+                "text": result.final_output,
             },
             *feedback_blocks,
         ]

openai-agents-sdk/listeners/views/issue_modal.py

@@ -67,11 +67,8 @@ def handle_issue_submission(ack: Ack, body: dict, client: WebClient, logger: Log
         feedback_blocks = create_feedback_block()
         response_blocks = [
             {
-                "type": "section",
-                "text": {
-                    "type": "mrkdwn",
-                    "text": result.final_output,
-                },
+                "type": "markdown",
+                "text": result.final_output,
             },
             *feedback_blocks,
         ]

pydantic-ai/agent/casey.py

@@ -20,7 +20,7 @@
 - Confident but honest when you don't know something
 
 ## Formatting Rules
-- Use Slack mrkdwn syntax: *bold*, _italic_, `code`, ```code blocks```, > blockquotes
+- Use standard Markdown syntax: **bold**, _italic_, `code`, ```code blocks```, > blockquotes
 - Use bullet points for multi-step instructions
 - Keep responses concise — aim for helpful, not verbose
 - When referencing ticket IDs or system names, use `inline code`