AI Memory Manager

Problem

In long conversations, project names, decisions, preferences, and deadlines are easy to lose in chat history. When work resumes, users often have to re-explain context, and assistants may miss constraints or accidentally store sensitive details. AI Memory Manager turns useful conversational details into searchable, manageable local records instead of leaving context trapped in the message thread.

How It Works

The skill listens in the background on every turn and evaluates whether information is worth saving based on the current sensitivity level: explicit project names, personal preferences, decisions, time commitments, standing instructions, or, in active mode, analytical conclusions. When a trigger matches, it asks the user first; only after approval does it append a structured entry to memory_bank.md, separated by --- and annotated with importance, tags, and a summary. Later, phrases such as “check memory” or “what happened with that project?” trigger retrieval, keyword matching, ranking, and display of up to the five most recent matches. It also supports weekly or monthly summaries, export, deletion, and full-clear management commands.

Boundaries

The skill relies on user confirmation and does not store passwords, API tokens, full phone numbers, detailed home addresses, or similar sensitive data. Privacy mode pauses new writes for the current conversation while existing memories remain searchable. It fits personal project follow-up, persistent collaboration preferences, and context recovery, but it is not designed for fully automatic memory, cross-device sync, or enterprise audit requirements.