Auto Memory Curation Skill
Overview
Silently analyze every message for important information and store it appropriately. Reduces manual memory management while building rich context over time.
How It Works
Trigger
- Runs on every message (silently)
- No user activation needed
Analysis Pipeline
Step 1: Filter Noise
Skip these message types:
- Greetings (hi, hello, hey)
- Thanks/thanking
- Acknowledgments (ok, sure, yes, yeah)
- Questions without context
- Single word responses
- Bot commands
Step 2: Categorize
For non-noise messages, categorize:
| Category |
What to look for |
Store in |
| Fact |
New info about user, projects, preferences |
MEMORY.md |
| Decision |
Choices made, conclusions reached |
memory/YYYY-MM-DD.md |
| Preference |
Likes, dislikes, style preferences |
USER.md |
| Idea |
Random thoughts, inspiration, concepts |
memory/topics/ideas.md |
| Learning |
Lessons, insights, discoveries |
memory/topics/lessons.md |
| Project |
Project updates, progress, blockers |
memory/projects/ |
| Goal |
Goals, targets, milestones |
memory/topics/goals.md |
| Error |
Mistakes, corrections to avoid |
memory/topics/anti-patterns.md |
| Commitment |
Promises I make, tasks to do |
tasks.md |
Step 3: Extract & Store
- Extract the key information
- Add timestamp reference
- Store in appropriate file
- Use append mode (never overwrite)
Guidelines
What to Capture
- New facts about Vini (name, preferences, goals)
- Project updates or decisions
- Ideas for future projects
- Learning insights
- Corrections (what doesn't work)
- Commitments I make
What to SKIP
- Passwords, secrets, API keys
- Trivial acknowledgments
- Basic confirmations
- Questions I'm asking
- Technical errors that are fixed
Quality Rules
- Be selective - Don't store everything
- Be concise - One sentence per memory
- Be contextual - Include enough info to understand later
- Be accurate - Don't paraphrase incorrectly
- Never duplicate - Check if already stored
Format
## [Category] - YYYY-MM-DD
- **[What]:** [Brief description]
*Context:* [Why it matters or relevant message]
Examples
User says:
"I prefer concise messages, no filler words"
Stored in USER.md:
Preference - 2026-03-06
- Communication: Prefers concise messages, no filler words
User says:
"Let's build a landing page for the consultancy inspired by Nexus AI"
Stored in memory/2026-03-06.md:
Decision - 2026-03-06
- Consultancy: Will use Nexus AI style for landing page
User says:
"I realized I work better with visual examples first, then theory"
Stored in MEMORY.md:
Learning - 2026-03-06
- Learning Style: Visual examples first, theory after
Testing
Periodically review stored memories to calibrate:
- Am I capturing too much noise?
- Am I missing important things?
- Are categories correct?
Adjust based on quality of accumulated memories.
Override
User can disable or adjust this skill at any time by saying:
- "Don't store that"
- "Clear recent memories"
- "Adjust memory curation"
1---2name: auto-memory-curation3description: Automatically analyzes messages for important information and stores it in the right memory files. Runs silently on every message. Filters noise, captures meaningful content. Use when you want passive memory building without manual "remember this" commands.4---56# Auto Memory Curation Skill78## Overview9Silently analyze every message for important information and store it appropriately. Reduces manual memory management while building rich context over time.1011## How It Works1213### Trigger14- Runs on every message (silently)15- No user activation needed1617### Analysis Pipeline1819**Step 1: Filter Noise**20Skip these message types:21- Greetings (hi, hello, hey)22- Thanks/thanking23- Acknowledgments (ok, sure, yes, yeah)24- Questions without context25- Single word responses26- Bot commands2728**Step 2: Categorize**29For non-noise messages, categorize:3031| Category | What to look for | Store in |32|----------|------------------|----------|33| **Fact** | New info about user, projects, preferences | MEMORY.md |34| **Decision** | Choices made, conclusions reached | memory/YYYY-MM-DD.md |35| **Preference** | Likes, dislikes, style preferences | USER.md |36| **Idea** | Random thoughts, inspiration, concepts | memory/topics/ideas.md |37| **Learning** | Lessons, insights, discoveries | memory/topics/lessons.md |38| **Project** | Project updates, progress, blockers | memory/projects/ |39| **Goal** | Goals, targets, milestones | memory/topics/goals.md |40| **Error** | Mistakes, corrections to avoid | memory/topics/anti-patterns.md |41| **Commitment** | Promises I make, tasks to do | tasks.md |4243**Step 3: Extract & Store**44- Extract the key information45- Add timestamp reference46- Store in appropriate file47- Use append mode (never overwrite)4849### Guidelines5051#### What to Capture52- New facts about Vini (name, preferences, goals)53- Project updates or decisions54- Ideas for future projects55- Learning insights56- Corrections (what doesn't work)57- Commitments I make5859#### What to SKIP60- Passwords, secrets, API keys61- Trivial acknowledgments62- Basic confirmations63- Questions I'm asking64- Technical errors that are fixed6566### Quality Rules67681. **Be selective** - Don't store everything692. **Be concise** - One sentence per memory703. **Be contextual** - Include enough info to understand later714. **Be accurate** - Don't paraphrase incorrectly725. **Never duplicate** - Check if already stored7374### Format7576```77## [Category] - YYYY-MM-DD7879- **[What]:** [Brief description]80 *Context:* [Why it matters or relevant message]81```8283### Examples8485**User says:**86> "I prefer concise messages, no filler words"8788**Stored in USER.md:**89> ## Preference - 2026-03-0690> - **Communication:** Prefers concise messages, no filler words9192---9394**User says:**95> "Let's build a landing page for the consultancy inspired by Nexus AI"9697**Stored in memory/2026-03-06.md:**98> ## Decision - 2026-03-0699> - **Consultancy:** Will use Nexus AI style for landing page100101---102103**User says:**104> "I realized I work better with visual examples first, then theory"105106**Stored in MEMORY.md:**107> ## Learning - 2026-03-06108> - **Learning Style:** Visual examples first, theory after109110---111112## Testing113114Periodically review stored memories to calibrate:115- Am I capturing too much noise?116- Am I missing important things?117- Are categories correct?118119Adjust based on quality of accumulated memories.120121## Override122123User can disable or adjust this skill at any time by saying:124- "Don't store that"125- "Clear recent memories"126- "Adjust memory curation"
dvcrn