memory-manager

Store and retrieve persistent project memory across chat sessions using TF-IDF search over JSONL storage.

Updated Apr 10, 2026
One-click install
npx skills add https://github.com/Exia-thd/Digital-Nervous --skill memory-manager-exia-thd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory-manager
Source: https://github.com/Exia-thd/Digital-Nervous/tree/main/skills/memory-manager
Command: npx skills add https://github.com/Exia-thd/Digital-Nervous --skill memory-manager-exia-thd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prevents loss of project context between sessions by providing a persistent, searchable memory that stores decisions, architecture notes, blockers, and session history so the orchestrator and skills avoid rediscovery and stay consistent across interactions.

Core Features & Use Cases

  • Persistent JSONL storage that is git-committed and namespaced per project for traceability and versioning.
  • TF-IDF search with cosine similarity and markdown-aware chunking to retrieve the most relevant memories within a token budget.
  • Safety and maintenance including secret redaction, .memignore support, CLI-driven ingestion/refresh/export, lifecycle hooks for automated adds at session start/turn close/phase gates, and value-weighted garbage collection to prune low-value entries.
  • Use Case: On session start the pipeline queries memory for recent decisions and blockers, injects the top results into the prompt to avoid repeating analysis, and records phase completions and gate decisions automatically.

Quick Start

Search project memory for authentication flow and return the top five relevant entries for context injection.

Frequently Asked Questions about memory-manager

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

FAQPage Schema
How do I persist project memory across chat sessions?

To persist project memory across chat sessions, store decisions, architecture notes, and blockers in a JSONL format that is namespaced per project and can be git-committed for traceability and versioning.

How does TF-IDF search retrieve relevant memories within a token budget?

TF-IDF search with cosine similarity retrieves relevant memories within a token budget by applying markdown-aware chunking to evaluate and return the most pertinent stored entries.

Can I use CLI-driven ingestion to export project memory and redact secrets?

Yes, CLI-driven ingestion supports export operations and automatically applies secret redaction to ensure sensitive information is scrubbed from the stored project memory.

What is the best way to prune low-value entries from persistent memory storage?

The best way to prune low-value entries from persistent memory storage is using value-weighted garbage collection, which automatically evaluates and removes entries to satisfy token and safety constraints.

When do I need to record gate decisions and phase transitions in project memory?

You need to record gate decisions and phase transitions during session start, before complex tasks, and at turn close using lifecycle hooks to prevent context loss across orchestrator pipelines.

Does persistent memory storage support .memignore for automated maintenance?

Yes, persistent memory storage supports .memignore files alongside lifecycle hooks and value-weighted garbage collection to automate maintenance and satisfy safety constraints.