AI Memory Context Management

Problem

After multi-turn conversations, AI assistants often cannot reliably retain user preferences, project progress, technical decisions, or daily context. memory-pro-v2 moves persistent information out of ephemeral session state and into local Markdown files, so later sessions can read, update, and recall it from clear paths.

How It Works

• Parse input: extract the information to remember, the intended category, and the date from natural language, such as work logs, project summaries, system preferences, or cross-team notes.
• Choose storage location: map the content to logs/, projects/, system/, or groups/ under .agents/data/memory/; high-frequency core information can go into the MEMORY.md summary file.
• Write and confirm: append or update the target file in standard Markdown with timestamps and category headings, then return the exact storage path to the user.
• Handle exceptions: ask for missing content when the memory is empty; automatically create missing category directories or initialize the MEMORY.md template when needed.

Scope And Caveats

This skill is suited to text-based context such as user preferences, project status, technical decisions, and daily logs. It is not appropriate for large binary files or code repositories. It requires filesystem read/write permission, and configuring Memory Recall rules in AGENTS.md is recommended so later sessions can reliably recall the relevant memory.