MindFlow - mindflow core prompts memorise
You are a world-class long-term memory management AI RAG Tool. Given a concise but detailed conversation context between a user and a conversational chatbot, summarise key points RELEVANT FOR FUTURE REFERENCE. DO NOT MEMORISE USELESS SHORT TERM POINTS.
EXAMPLES OF RELEVANT FUTURE REFERENCES AND SUMMARIES:
1. User loves winter season
2. User has friends called Tyler, Callum, and Sam
3. User has an internship interview for PwC tomorrow (12 October 2024)
4. User lives in Christchurch, NZ
5. User is studying Bachelor's of Software Engineering at University of Canterbury
6. User has a goal to study overseas
7. User loves the subject maths
**IMPORTANT: RETURN ONLY IN THE FOLLOWING JSON FORMAT**
**IMPORTANT: FIELDS ARE ONLY ALLOWED TO BE str, int, float or bool**
**IMPORTANT: DO NOT REPEAT THE SAME MEMORY IF PREVIOUSLY ALREADY MENTIONED**
{
"memory": "User wants to go to restaurant called 'Big Chicken' with John",
"metadata": {
"involved": "User, John"
}
}
**IMPORTANT: IF THERE IS NOTHING RELEVANT, SIMPLY RETURN:**
{}
Instructions flagged against the user
D3 · Privacy & Data Protection
“User has friends called Tyler, Callum, and Sam”
The prompt instructs the system to store personal information about users (friends' names, location, education, goals, interview details) without any mention of user consent, data retention policies, transparency about what is being stored, or data minimization principles. The examples explicitly encourage memorizing PII such as names of friends, home city, employer names, and educational institution. There is no instruction to inform the user about what is being memorized or to seek permission.
D3 · Privacy & Data Protection
“User has an internship interview for PwC tomorrow (12 October 2024)”
The prompt instructs the system to store personal information about users (friends' names, location, education, goals, interview details) without any mention of user consent, data retention policies, transparency about what is being stored, or data minimization principles. The examples explicitly encourage memorizing PII such as names of friends, home city, employer names, and educational institution. There is no instruction to inform the user about what is being memorized or to seek permission.
D3 · Privacy & Data Protection
“User lives in Christchurch, NZ”
The prompt instructs the system to store personal information about users (friends' names, location, education, goals, interview details) without any mention of user consent, data retention policies, transparency about what is being stored, or data minimization principles. The examples explicitly encourage memorizing PII such as names of friends, home city, employer names, and educational institution. There is no instruction to inform the user about what is being memorized or to seek permission.
D3 · Privacy & Data Protection
“User is studying Bachelor's of Software Engineering at University of Canterbury”
The prompt instructs the system to store personal information about users (friends' names, location, education, goals, interview details) without any mention of user consent, data retention policies, transparency about what is being stored, or data minimization principles. The examples explicitly encourage memorizing PII such as names of friends, home city, employer names, and educational institution. There is no instruction to inform the user about what is being memorized or to seek permission.
D3 · Privacy & Data Protection
“"memory": "User wants to go to restaurant called 'Big Chicken' with John",”
The prompt instructs the system to store personal information about users (friends' names, location, education, goals, interview details) without any mention of user consent, data retention policies, transparency about what is being stored, or data minimization principles. The examples explicitly encourage memorizing PII such as names of friends, home city, employer names, and educational institution. There is no instruction to inform the user about what is being memorized or to seek permission.
D5 · User Agency & Manipulation Prevention
“Given a concise but detailed conversation context between a user and a conversational chatbot, summarise key points RELEVANT FOR FUTURE REFERENCE.”
The system operates as a background memory extraction tool with no indication that users are informed about or can control what is being memorized. There is no mechanism for user consent, review, or deletion of stored memories. The user has no agency over what the system decides is 'relevant for future reference.'
D5 · User Agency & Manipulation Prevention
“Given a concise but detailed conversation context between a user and a conversational chatbot, summarise key points RELEVANT FOR FUTURE REFERENCE. DO NOT MEMORISE USELESS SHORT TERM POINTS.”
The system operates as a background memory extraction tool with no indication that users are informed about or can control what is being memorized. There is no mechanism for user consent, review, or deletion of stored memories. The user has no agency over what the system decides is 'relevant for future reference.'