Caduceus Auto-Save

Automatically detect when person-related information is mentioned and save it to Core Memory for future reference.

When This Activates

This skill activates when you detect ANY of these patterns in user messages:

•A person's name with their role/title ("Sarah is the CEO", "my manager John")
•Contact information (email, phone, WhatsApp, LinkedIn)
•Organizational relationships ("works at", "in charge of", "leads the")
•Interpersonal relationships ("reports to", "works with", "is married to")
•Professional context ("specializes in", "expert in", "handles")

Extraction Schema

When person information is detected, extract:

Field	Description	Example
name	Full name	"Isa Derteano", "Dr. James Chen"
role	Job title/position	"Head of Marketing", "CTO"
organization	Company/team	"ALQVEMENTS", "Google Cloud Team"
contact	Email, phone, WhatsApp, social	"+34 722 108 179", "isa@company.com"
relationships	Connections to others	"reports to Ormus", "manages design team"
responsibilities	What they handle	"oversees all MANA marketing"
expertise	Skills/domains	"brand strategy", "Python", "UX design"
notes	Context, how you met	"Met at conference", "Friend of Diego"

How to Save

Use the Core Memory tools:

•

Initialize session (if not already done):

code

mcp__core-memory__initialize_conversation_session

•
Save the person info:
code
```
mcp__core-memory__memory_ingest
```

Format the memory entry as:

code

PERSON: [Full Name]
Role: [Title] at [Organization]
Contact: [email/phone/WhatsApp]
Relationships: [connections to other people]
Responsibilities: [what they handle]
Expertise: [skills/domains]
Notes: [context, how met, relevant background]
Source: [conversation where this was learned]

Behavior Rules

•
Always confirm after saving: "Saved [Name] to Caduceus: [brief summary]"
•
Update existing entries - If person already exists, add new info rather than duplicate
•
Bi-directional relationships - If "A reports to B", also note that "B manages A"
•
Preserve context - Include when/how this information was learned
•
Don't assume - Only save explicitly stated information, don't infer

Examples

Example 1: Role Assignment

User says: "Isa Derteano is the person in charge of marketing for ALQVEMENTS"

Extract & Save:

code

PERSON: Isa Derteano (Isabel Derteano)
Role: Head of Marketing at ALQVEMENTS
Responsibilities: Oversees marketing for all ALQVEMENTS brands including MANA Supplements
Contact: WhatsApp +34 722 108 179

Confirm: "Saved Isa Derteano to Caduceus: Head of Marketing at ALQVEMENTS"

Example 2: Contact Info

User says: "My accountant is Maria Lopez, her email is maria@accounting.com and she handles all ALQVEMENTS finances"

Extract & Save:

code

PERSON: Maria Lopez
Role: Accountant
Organization: External / ALQVEMENTS
Contact: maria@accounting.com
Responsibilities: Handles all ALQVEMENTS finances
Relationships: Works with Ormus on financial matters

Example 3: Team Structure

User says: "Diego manages the dev team. His direct reports are Ana (frontend), Carlos (backend), and Lucia (design)"

Extract & Save (multiple entries):

Entry 1:

code

PERSON: Diego
Role: Dev Team Manager
Relationships: Manages Ana, Carlos, Lucia

Entry 2-4: Ana, Carlos, Lucia with "Reports to Diego" relationship

Example 4: Professional Network

User says: "I met Alex at the conference, he's a VC at Sequoia focusing on wellness startups. Might be good for MANA funding."

Extract & Save:

code

PERSON: Alex
Role: Venture Capitalist at Sequoia
Expertise: Wellness startups, funding
Notes: Met at conference. Potential investor for MANA.

What NOT to Save

•Casual mentions without descriptive info ("I talked to John today")
•Public figures unless personally connected
•Speculative relationships ("I think she might work at...")
•Sensitive personal info not voluntarily shared

Output Format

After saving, always confirm:

Saved to Caduceus: [Name] - [Role] at [Organization] [Brief summary of key info saved]