memory-manager

Store project decisions and progress as git-committed JSONL records.

47|11|Updated Mar 6, 2026
One-click install
npx skills add https://github.com/buiphucminhtam/forgewright --skill memory-manager-buiphucminhtam
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory-manager
Source: https://github.com/buiphucminhtam/forgewright/tree/main/skills/memory-manager
Command: npx skills add https://github.com/buiphucminhtam/forgewright --skill memory-manager-buiphucminhtam

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Persistent project memory is often lost between sessions; this skill provides a git-committed JSONL store with TF-IDF search, markdown-aware chunking, and a value-weighted GC to preserve decisions, architecture, blockers, and task status across sessions.

Core Features & Use Cases

  • Persistent memory store: git-committed JSONL that tracks decisions, architecture, blockers, tasks, and session history.
  • Fast retrieval: TF-IDF search with markdown-aware chunking to surface relevant context quickly.
  • Lifecycle integration: integrates with Forgewright pipeline to maintain cross-session continuity and improve decision traceability.
  • Garbage collection: value-weighted pruning to keep memories focused and manageable.

Quick Start

Ingest your project data and start memory retrieval for the current session.

Frequently Asked Questions about memory-manager

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

FAQPage Schema
How do I keep project memory persistent across sessions?

Persistent project memory is maintained across sessions using git-committed JSONL records that preserve decisions, architecture, blockers, and tasks. The value-weighted garbage collection prunes low-value data to keep the store focused and manageable.

What is TF-IDF search for retrieving project context?

TF-IDF search retrieves relevant project context by applying term frequency analysis to markdown-aware chunks. This mechanism surfaces the most pertinent decisions and historical session data quickly without scanning entire documents.

How do I store cross-session decisions and architecture in a git workflow?

Cross-session decisions and architecture are stored as git-committed JSONL records, tracking decisions, architecture, blockers, and tasks. This integrates with the Forgewright pipeline to maintain continuity and improve decision traceability.

Does memory-manager work without external dependencies?

Yes, the memory management skill operates with no external dependencies. It relies entirely on git-committed JSONL files and built-in TF-IDF search, requiring no additional packages or isolated database services to function.

How does value-weighted garbage collection prune low-value memory?

Value-weighted garbage collection prunes low-value memory by evaluating stored JSONL records and removing entries that lack ongoing relevance. This keeps the persistent project memory focused, searchable, and manageable over time.