Overview

Memory Analyzer scans conversation history to extract user preferences and feedback. It then updates long-term memory files (MEMORY.md, AGENTS.md, USER.md, IDENTITY.md, SOUL.md) to keep a personalized interaction context across sessions.

How This Skill Works

It uses the default google/gemini-3-flash-preview model to read sessions/, detect preferences and feedback patterns, and automatically write insights into the memory files, enabling continuous, personalized behavior across interactions.

When to Use It

• You want to capture evolving user preferences directly from chat history
• After explicit feedback or format preferences to refine behavior
• When building persistent agent rules and identity across sessions
• To keep long-term memory files in sync (MEMORY.md, USER.md, etc.)
• During onboarding or role clarification to lock in personality and goals

Quick Start

1. Step 1: Run the memory-analyzer skill with google/gemini-3-flash-preview (default).
2. Step 2: Let it parse sessions/ and extract user preferences and feedback.
3. Step 3: Review MEMORY.md, AGENTS.md, USER.md, IDENTITY.md, SOUL.md and verify updates.

Best Practices

• Use explicit feedback phrases to trigger updates
• Review memory changes regularly to avoid drift
• Keep memory files human-readable with concise rules
• Respect privacy and opt-in constraints
• Test with edge cases to ensure correct extraction and updates

Example Use Cases

• User states a preference for concise responses; MEMORY/USER updates reflect this to shorten future replies.
• Frequent feedback on formatting leads to AGENTS.md adjustments for response structure.
• Role clarification updates IDENTITY.md to align tone and capabilities.
• Consistent tone feedback updates SOUL.md to adjust personality settings.
• Long-term habits are stored in MEMORY.md to influence recommendations across sessions.

Frequently Asked Questions