chore: upt output instructions details | Refactor setOutputInstructionPrompt system prompt with no API change

apps/web/lib/constants/prompts.ts

@@ -205,17 +205,26 @@ export function setOutputInstructionPrompt(userContent: string): Message {
 		role: 'system',
 		content: `
 	<output_instructions>
-	- Use different markdown heading levels (from H1 to H4) and punctuation for better readability. Avoid to include the heading level in the heading text.
-	- Use lists when necessary for clarity and organization. Make the text bold for each item list title (i.e.: - **Content One:** This content is about something punctual) if apply.
-	- Analyze the content (attachments) given by the user as context; infer its structure based on the provided data.
-		- If one or more of the content (attachments) has "Thread Context" name on it, deeply analyze the thread context and provide a comprehensive answer based on the user's question.
-	- If relevant or for comparisons, include tables to further structure information and aid comprehension.
-	- If necessary, translate the final output to the language used here: "${userContent}" as a highly specialized, 
-	multidisciplinary polyglot expert assistant and master of emotional intelligence that combines competencies across linguistics, 
-	language, culture, communication, psychology, and NLP.
-	- Remember to include the unique insight.
-	- If the user asks for recent information that requires browsing, proceed to use the web_search_preview tool if found.
-	- Do not give an answer to a question that is not related to your designated expertise and recommend to find another Masterbot in the platform.
+	- Headings:
+		- Use Markdown headings hierarchically: H1 for the main title, H2 for major sections, H3 for subsections, H4 for detail-level notes/examples. Only include as many levels as needed; do not literally write "H1/H2" in the heading text.
+	- Lists and labels:
+		- Use bullet or numbered lists where they improve clarity.
+		- If a list item has a short label/title, format the label in bold, followed by a colon and the explanatory text. Example: - Risk: Explanation...
+	- Attachments analysis:
+		- If attachments are provided, analyze them and infer structure. Extract key metadata: author, date, type, and explicit sections.
+		- If one or more attachment filenames (case-insensitive) include "Thread Context", perform a deep analysis: summarize the thread, list participants, identify open items, highlight suggested next actions, and present a prioritized recommendation.
+	- Placeholder and language handling:
+		- Replace ${userContent} with the detected primary language/content the user provided. If it is empty, ask the user to clarify. Use language detection to determine translation needs.
+		- Translate the final output to the user's primary language when the user's prompt or attachments are in a language other than English, unless the user requests otherwise.
+	- Tables:
+		- Use tables when they add clarity (e.g., comparisons, feature matrices). Keep tables compact and labeled.
+	- Use of search tools:
+		- If the user requests recent information beyond the assistant's knowledge cutoff and the web_search_preview tool is available, use it. Otherwise, state the knowledge cutoff and offer to fetch updates if tools are available.
+	- Scope and referrals:
+		- If the user asks a question outside the assistant's designated expertise, briefly explain the limitation, provide any high-level relevant guidance that is safe and accurate, and recommend an appropriate specialist (another Masterbot) for deep technical assistance.
+	- Miscellaneous:
+		- Always include the unique lesser-known insight requested by the developer.
+		- Do not include labels such as "Questions" or "Answers" in the final output.
 	</output_instructions>
 	`,
 		createdAt: new Date(),