Can I manually edit or delete memory files?

Yes, because memories are stored as standard Markdown files, you can manually update, reorganize, or delete them using any text editor or file manager.

How does Claude find relevant memories?

Claude uses a summary-first approach, scanning the frontmatter of memory files using ripgrep to quickly identify relevant information before reading the full content.

What is the Agent Memory skill?

It is a specialized capability for Claude Code that allows the AI to save and retrieve important information in a local directory, enabling it to remember context across different chat sessions.

Where are memories stored?

Memories are stored as Markdown files with YAML frontmatter in the .claude/skills/agent-memory/memories/ directory within your project.

Is the memory shared across different projects?

No, the memory is project-specific and resides within the project's local directory, ensuring that the stored knowledge remains relevant to that specific codebase.

Persistent Agent Memory

Name: Persistent Agent Memory
Author: yamadashy

byyamadashy

•

21,364

生产力与工作流

Enables Claude to store and recall long-term knowledge, research findings, and project context across multiple conversations.

关于

Persistent Agent Memory provides a structured, file-based storage system that allows Claude to retain critical information between sessions. By saving architectural decisions, complex research results, and work-in-progress notes to a dedicated directory, this skill ensures that valuable context is never lost. It uses a summary-first search approach with ripgrep to efficiently retrieve relevant memories, helping developers maintain continuity and avoid repeating technical investigations throughout the lifecycle of a project.

主要功能

21,364 GitHub stars
Summary-first search workflow for efficient information retrieval
Proactive knowledge capture of architectural decisions and patterns
Standardized Markdown format with YAML frontmatter for status tracking
Organized folder structure with support for custom categories and tags
Persistent file-based storage for long-term context retention

使用场景

Documenting non-obvious codebase 'gotchas' or performance patterns
Maintaining a persistent log of architectural decisions and their rationales
Resuming complex coding tasks after a multi-day break with full context

关于

主要功能

21,364 GitHub stars
Summary-first search workflow for efficient information retrieval
Proactive knowledge capture of architectural decisions and patterns
Standardized Markdown format with YAML frontmatter for status tracking
Organized folder structure with support for custom categories and tags
Persistent file-based storage for long-term context retention

使用场景

Documenting non-obvious codebase 'gotchas' or performance patterns
Maintaining a persistent log of architectural decisions and their rationales
Resuming complex coding tasks after a multi-day break with full context