memory-manager

To prevent agent context window overflow, you need memory pruning strategies that remove stale memories and consolidate important facts across short- and long-term stores. This Skill manages memory growth and applies importance scoring to keep working context aligned with current conversations.

Importance scoring for agent memory pruning evaluates and prioritizes memories based on their relevance and decay over time. This mechanism monitors memory growth, removes low-importance entries to reduce noise, and ensures high-value facts are retained for retrieval.

Optimizing conversation history archiving requires consolidating important information into longer-lived knowledge stores. This Skill manages memory create, retrieve, search, and update operations to keep critical facts available across sessions while pruning outdated context.

Yes, semantic search is supported alongside memory management to retrieve relevant agent context. The Skill applies deterministic pruning strategies and decay-based prioritization to ensure semantic search queries return high-importance memories without context-window bloat.

You should use deterministic pruning strategies for agent runtime memory when low-importance or stale entries degrade retrieval quality and cause context-window bloat. Applying these strategies during active conversations keeps agent memory fast, relevant, and safe.

Your agent retrieves outdated facts because working context is not properly aligned with current conversations. Applying decay-based prioritization and pruning stale memories ensures that recent, high-importance facts are consolidated and retrieved accurately.