How does the skill prevent duplicate information?

The skill utilizes a deduplication flag during storage that checks for a 0.9 similarity threshold, preventing near-duplicate entries from cluttering search results.

What are memory tiers in the Memory Workflow skill?

Memory tiers determine retention and retrieval priority: 'Hot' memories are critical facts always loaded at startup, 'Warm' memories are retrieved via search, and 'Cold' memories are archived historical notes.

What is the difference between a memory and a decision?

While memories store general information, architectural choices should be handled via the decision_record tool (linked to the decision-tracking skill) for better specialized tracking.

How do I ensure my project context is loaded correctly?

Always follow the startup sequence: first call project_context to load hot-tier memories, then session_start to begin tracking the current work session.

Should I use this skill to store API keys or passwords?

No. The system includes a blocklist that automatically rejects credentials; sensitive information should always be handled by a dedicated secrets manager.

Memory Workflow

Name: Memory Workflow
Author: memoryrelay

bymemoryrelay

0•

生产力与工作流

Provides persistent, cross-session memory management to maintain project context and architectural state throughout the development lifecycle.

The Memory Workflow skill enables Claude to maintain a long-term memory across different chat sessions, ensuring that crucial project facts, architectural decisions, and coding patterns are never lost. It facilitates a structured startup sequence to load project context, allows for semantic search of past information, and utilizes a tiered storage system—Hot, Warm, and Cold—to prioritize critical information. By implementing mandatory deduplication and metadata tagging, it prevents context clutter and ensures that Claude always has access to the most relevant historical data for any given task.

主要功能

010 GitHub stars

02Metadata-driven categorization for structured filtering of technical facts and decisions

03Session tracking with automated summaries for chronological record-keeping

04Semantic search and retrieval for finding information across historical sessions

05Automatic deduplication with similarity thresholds to maintain data integrity

06Tiered memory storage (Hot, Warm, Cold) for optimized context loading

使用场景

01Retrieving specific technical decisions or rationale made in previous coding sessions

02Maintaining architectural context and project facts across long-running development cycles

03Building a persistent knowledge base of project-specific conventions and established patterns

主要功能

010 GitHub stars

02Metadata-driven categorization for structured filtering of technical facts and decisions

03Session tracking with automated summaries for chronological record-keeping

04Semantic search and retrieval for finding information across historical sessions

05Automatic deduplication with similarity thresholds to maintain data integrity

06Tiered memory storage (Hot, Warm, Cold) for optimized context loading

使用场景

01Retrieving specific technical decisions or rationale made in previous coding sessions

02Maintaining architectural context and project facts across long-running development cycles

03Building a persistent knowledge base of project-specific conventions and established patterns