Can I manually trigger the learning process?

Yes, you can use triggers like --project-store or --learn, and the skill also automatically activates when it detects significant project insights or developer frustration.

Does this skill support vector databases?

Yes, it supports optional dual-writing to Qdrant MCP servers, allowing for semantic search capabilities alongside the primary file-based storage.

Why does the skill track failed development trajectories?

Tracking failures is critical for AI performance; it prevents Claude from suggesting solutions or approaches that have already been proven ineffective in your specific environment.

How does Project Memory Store differ from universal AI knowledge?

Universal knowledge covers general coding practices, while Project Memory Store focuses exclusively on the unique patterns, decisions, and specific logic of your particular codebase.

Where are the project memories stored?

Memories are stored locally in your repository under the .claude/skills/project-memory-store/ directory in a human-readable Markdown format.

Project Memory Store

Name: Project Memory Store
Author: hongbietcode

byhongbietcode

•

生産性とワークフロー

Captures and organizes project-specific insights, architectural decisions, and failure patterns into a persistent file-based memory system.

概要

Project Memory Store is a specialized Claude Code skill designed to bridge the context gap in long-term software development. It enables Claude to extract, classify, and store episodic events, procedural workflows, and semantic patterns directly within your repository. By documenting not just what worked, but also the 'failure trajectories' and frustrations encountered during development, this skill ensures that Claude evolves alongside your codebase. It utilizes a structured file-based hierarchy and optional vector search integration to provide a 'living documentation' layer that prevents repetitive mistakes and maintains architectural consistency across sessions.

主な機能

Categorized memory storage for episodic events, workflows, and project patterns
Dual-write support for local Markdown files and Qdrant vector databases
2 GitHub stars
Frustration-triggered learning to capture and avoid recurring failure patterns
Clean hierarchical organization with progressive disclosure for optimized retrieval
Automated memory consolidation to merge duplicates and generalize new insights

ユースケース

Documenting unique architectural decisions and domain-specific logic for long-term maintenance
Standardizing custom deployment or testing workflows that are unique to a specific codebase
Building a project-specific 'avoid-list' based on past debugging failures and anti-patterns

概要

主な機能

Categorized memory storage for episodic events, workflows, and project patterns
Dual-write support for local Markdown files and Qdrant vector databases
2 GitHub stars
Frustration-triggered learning to capture and avoid recurring failure patterns
Clean hierarchical organization with progressive disclosure for optimized retrieval
Automated memory consolidation to merge duplicates and generalize new insights

ユースケース

Documenting unique architectural decisions and domain-specific logic for long-term maintenance
Standardizing custom deployment or testing workflows that are unique to a specific codebase
Building a project-specific 'avoid-list' based on past debugging failures and anti-patterns