reasoning-bank

What problem does it solve?

This Skill solves noisy, flat memory retrieval by ranking memories by both confidence and recency, and by letting you deliberately reinforce what proved useful.

Core Features & Use Cases

• Confidence-weighted retrieval: Searches memory directories and ranks matches using confidence × recency decay so the most relevant memories surface first.
• Deliberate boosting loop: Updates a memory’s frontmatter (used_count, last_used, and confidence) with a controlled diminishing-returns curve.
• Pruning-friendly maintenance signal: Flags older, low-confidence entries for eventual decay/pruning by keeping reinforcement and decay semantics aligned with the existing maintenance system.

Quick Start

Run /reasoning-bank retrieve <query> to get the top-ranked memories relevant to your question.