Store and retrieve user context via embedding-based semantic similarity search.
npx skills add https://github.com/harperreed/dotfiles --skill memory-harperreed
Please help me install this Agent Skill. Skill: memory Source: https://github.com/harperreed/dotfiles/tree/main/.claude/skills-archive/memory Command: npx skills add https://github.com/harperreed/dotfiles --skill memory-harperreed
Memory helps you store useful information so it can be recalled later without rereading past conversations or notes.
Use the memory skill to save a preference by asking: memory add "User prefers TypeScript over JavaScript for new projects" with tags preferences and programming.
High-intent search queries and answers about installing and using this skill.
To persist user context for cross-session recall, you store preferences and decisions using semantic memory tools that save information as embeddings. This allows future sessions to retrieve relevant knowledge by meaning without rereading past conversations.
Semantic similarity search retrieves stored memory by converting queries into embeddings and comparing them against saved context vectors. This matches items by underlying meaning rather than requiring exact keyword overlap, returning the most relevant snippets.
The best way to log project-specific context is using the memory skill's extraction workflows to save important details as embeddings. This ensures long-term knowledge persistence by capturing project-specific information for future semantic retrieval.
Yes, you need MCP memory tools to use this semantic memory skill. The skill requires MCP tools specifically for storing vectors, searching by query meaning, listing recent entries, deleting items, and returning relevant context snippets.
Yes, you can list recent memory entries and delete stored items when they are no longer accurate. The memory management workflow includes list and delete operations to help maintain relevant context and remove outdated information.
You should not use embedding-based semantic memory when exact keyword matching is required, or when the overhead of managing vector storage outweighs the benefit of semantic recall for short, highly transient conversations that do not require long-term persistence.
ThomazRossitoCreate, manage, and query vector indexes in Databricks Vector Search.
cryptopafiSearch and analyze session JSONL logs by date, session, or keyword.
romiluz13Review persistent agent memory files and database entries using a five-outcome classification rubric.
FalkorDBCreate and query HNSW vector indexes in FalkorDB graph databases.
