How does Claude handle unknown terms or shorthand?

Claude follows a tiered lookup: it checks CLAUDE.md first, then searches memory/glossary.md. If the term is still not found, it will ask the user for clarification and then save that definition for future use.

What is the difference between CLAUDE.md and the memory directory?

CLAUDE.md acts as a 'hot cache' for your most frequent contacts and terms (roughly the top 30), while the memory/ directory serves as the full, searchable knowledge base for everything else.

Can I use this for multiple teams and projects?

Yes. The system includes dedicated subdirectories for projects and context, allowing you to maintain detailed histories, stakeholder lists, and specific tool configurations for any number of initiatives.

How do I initialize my memory for the first time?

You can use the /productivity:start command to bootstrap the system. It scans your environment—including chat history and documents—to extract key people and projects automatically.

Memory Management

Name: Memory Management
Author: anthropics

byanthropics

•

8,000

•

생산성 및 워크플로

Builds a persistent, two-tier knowledge base to help Claude understand internal company language, shorthand, and project context.

Memory Management transforms Claude into a high-context workplace collaborator by implementing a structured system for storing and retrieving domain-specific knowledge. It utilizes a tiered architecture consisting of a 'hot cache' in CLAUDE.md for immediate access to frequent terms and a comprehensive directory for deep context on people, projects, and internal glossaries. This allows Claude to decode shorthand and acronyms instantly, ensuring it operates with the same institutional knowledge as a long-term team member while maintaining a lean and efficient working memory.

주요 기능

01Automated bootstrapping to initialize memory from existing chats, calendars, and documents

02Rich profile management for team members, stakeholders, and organizational roles

038,000 GitHub stars

04Internal language decoding for acronyms, nicknames, and project codenames

05Two-tier storage system utilizing CLAUDE.md for hot cache and a memory/ directory for scale

06Progressive disclosure logic to keep working memory lean while supporting unlimited deep knowledge

사용 사례

01Maintaining a long-term repository of team preferences and company-specific operational processes

02Onboarding Claude to complex projects with specific internal terminology and stakeholders

03Managing high-volume task requests that use team shorthand and obscure project names

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add anthropics/knowledge-work-plugins memory-management

For use in Claude.ai and ChatGPT

Download Skill