Does this skill save my entire codebase?

No, it saves structured metadata including absolute file paths, status markers, and logic summaries. This keeps the memory footprint small while providing enough context to resume work.

Can I add my own notes to the memory snapshot?

Yes, the skill supports a --description argument for custom session summaries and a --tags argument for easier categorization and searching later.

Why does it use absolute paths instead of relative ones?

Absolute paths ensure that when you resume a session, the AI can find the exact files regardless of which directory the new terminal session was started in.

What is a Recovery ID?

A Recovery ID is a unique timestamped identifier (e.g., CMEM-20231027-103000) generated after a successful compact operation, allowing you to re-import that specific state in a future session.

How do I resume a saved session?

You can restore a session by using the command 'Please import memory ' or by utilizing the MCP core_memory export tool with your specific Recovery ID.

Memory Compact

Name: Memory Compact
Author: catlog22

bycatlog22

•

1,371

•

生産性とワークフロー

Compresses active session context into structured, recoverable snapshots to ensure seamless task resumption and state persistence.

Memory Compact is a high-fidelity workflow utility designed to bridge the gap between AI coding sessions by capturing the essential 'DNA' of a project's current state. It extracts critical metadata—including verbatim execution plans, absolute file paths, architectural decisions, and pending tasks—into a structured format optimized for session recovery. By storing these snapshots in persistent memory via MCP tools, it eliminates the need for redundant re-exploration, preserves the developer's train of thought during complex debugging, and ensures that the AI agent always has a 'North Star' to follow when resuming work.

主な機能

01Saves session snapshots to persistent storage using MCP core_memory integration

02Identifies modified 'working' files vs. read-only 'reference' files for immediate context

03Automatically resolves absolute file paths for reliable cross-session navigation

04Captures complete execution plans verbatim to prevent re-planning loops

05Extracts architectural decisions and reasoning to maintain design consistency

061,371 GitHub stars

ユースケース

01Maintaining a persistent 'North Star' objective across long-running development workflows

02Resuming a complex multi-file refactoring after a session timeout or context window reset

03Handing off a detailed debugging state including hypotheses and known issues

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add catlog22/claude-code-workflow memory-compact

For use in Claude.ai and ChatGPT

Download Skill