Remember

A persistent memory skill for AI coding assistants that stores user preferences, project conventions, and workflow decisions across sessions, ensuring consistent behavior without repeating instructions.

What Is This?

Overview

This skill enables AI assistants to remember context between conversations. It saves key information like coding style preferences, project architecture decisions, preferred tools, naming conventions, and recurring instructions into persistent memory files. When a new session begins, the assistant loads these memories and applies them automatically. This eliminates the friction of re-explaining your setup every time you start working with the AI.

Who Should Use This

Perfect for developers who work with AI coding assistants regularly and find themselves repeating the same preferences. Especially useful for teams with established conventions and developers working on long running projects.

Why Use It?

Problems It Solves

Without persistent memory, every AI conversation starts from zero. Developers must re-specify their test framework, explain project structure, and correct the same mistakes repeatedly. This wastes time and creates inconsistency across sessions.

Core Highlights

• Cross Session Persistence retains information between separate conversations
• Automatic Application loads saved preferences without manual prompting
• Organized Storage keeps memories categorized by topic for easy management
• User Controlled developers decide what gets saved and can update or remove entries
• Project Scoped supports different memory sets for different projects

How to Use It?

Basic Usage

Tell the assistant to remember something and it will persist that information for future sessions.

"Always use pnpm instead of npm for this project"
"We use vitest not jest for testing"
"Our API base URL is /api/v2"
"Never auto-commit, always ask first"

Real-World Examples

Full Stack Project Setup

A developer working on a Next.js project with specific conventions told the assistant their preferences once. The memory skill saved them, and every subsequent session automatically followed the same patterns.

## Architecture
- Next.js 14 App Router
- Prisma ORM with PostgreSQL
- Tailwind CSS with shadcn/ui components

## Conventions
- File naming: kebab-case for files, PascalCase for components
- API routes return { data, error } shape
- All database queries go through service layer in src/services/

## Preferences
- Use server components by default, add "use client" only when needed
- Prefer named exports over default exports
- Run lint and type check before suggesting PR creation

Advanced Tips

Organize memories into separate topic files linked from a main index. Keep the primary file concise and use linked files for details. Periodically prune outdated memories to prevent conflicts.

When to Use It?

Use Cases

• Coding Style Preferences save formatting, naming, and structural conventions
• Tool Configuration remember which package manager, test runner, and linter to use
• Project Architecture store decisions about folder structure, patterns, and frameworks
• Workflow Rules persist instructions like commit message format or review requirements
• Team Conventions share consistent assistant behavior across team members

Related Topics

When working with persistent memory, these prompts activate the skill:

• "Remember that I always use TypeScript strict mode"
• "Save this preference for future sessions"
• "Do not forget that we use Prettier with tabs"
• "What do you remember about this project"

Important Notes

Requirements

• Requires a file system location for persistent storage, typically a dotfile directory
• Works with any AI assistant framework that supports loading context from files
• Benefits from version control for memory files to track changes over time
• Memory files should be kept concise to avoid exceeding context window limits

Usage Recommendations

Do:

• Be specific when saving memories so the assistant can apply them precisely
• Review memories periodically to remove outdated or conflicting entries
• Organize by topic rather than chronologically for easier maintenance
• Verify memories against project docs before trusting them as authoritative

Don't:

• Save session specific context like current task details or temporary state
• Store sensitive information like API keys or passwords in memory files
• Let memories accumulate unchecked as contradictions will confuse the assistant
• Duplicate instructions that already exist in project configuration files

Limitations

• Memory files have size limits to fit within AI context windows, so prioritize what gets saved
• Cannot automatically detect when saved information becomes outdated
• Memories are text based and cannot store complex state or binary data
• Different AI tools may use different memory formats, limiting portability across platforms