short-term-memory

Maintain per-project working context in a current-context.md file.

21|1|Updated Feb 27, 2026
One-click install
npx skills add https://github.com/Malaiac/claude --skill short-term-memory
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: short-term-memory
Source: https://github.com/Malaiac/claude/tree/main/skills/short-term-memory
Command: npx skills add https://github.com/Malaiac/claude --skill short-term-memory

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Claude Code loses recent momentum, decisions, and the next steps after a context compaction or session resume.

Core Features & Use Cases

  • Per-project current-context.md storage that is read before every response and updated after
  • Timestamped entries with automatic archiving when the log grows
  • Keeps recent decisions, blockers, and next steps across compactions and sessions

Quick Start

Read current-context.md before every response and update it with a timestamp after each response.

Frequently Asked Questions about short-term-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I preserve context across Claude Code session compactions?

To preserve context across session compactions, use a per-project current-context.md file that stores recent decisions and next steps. Claude reads this file before every response and updates it with a timestamp after each interaction, ensuring momentum survives compactions.

What is per-project memory persistence for coding workflows?

Per-project memory persistence is the practice of maintaining a current-context.md file in your project directory. This file is read before responses and updated after interactions, keeping recent decisions, blockers, and next steps available across sessions and compactions.

How do I set up a current-context file for my project?

Place a current-context.md file directly in your project directory. Entries should use the YYYYMMDD-HHMMSS timestamp format and remain concise. The file is read before every response and updated after each interaction, with older entries archived when the log grows.

Does context persistence work across any project directory?

Yes, this context persistence approach applies across any project. It requires placing the current-context.md file in the project directory and keeping entries concise while preserving recent decisions and next steps, making it platform-agnostic for workflow automation.

What happens when the current-context log grows too large?

When the current-context log grows too large, older entries are automatically archived. This keeps the file concise and ensures only recent decisions, blockers, and next steps are preserved for reading before every response during session resumes.

Why does Claude Code lose momentum after a session resume?

Claude Code loses momentum after a session resume because recent decisions and next steps are not retained by default. Maintaining a current-context.md file that survives compactions and resumes solves this by preserving working context per-project.