What problem does it solve? The Azoth memory system reliably writes M3 episodes and M2 patterns but has no automatic read-back path, so every session starts cold regardless of accumulated prior context. This Skill closes the write-only memory sink by surfacing relevant past lessons before planning begins. ## Core Features & Use Cases - Tag-Based Scoring: Extracts 3-5 tags from the current goal and scores M3 episodes (tag overlap, recency, reinforcement count) and M2 patterns (trigger match, reinforcements) to rank candidates. - Conflict Handling: Detects contradictions, stale lessons, and superseded items, flagging them explicitly instead of collapsing them into false consensus. - Structured Recall Output: Surfaces the top 1-3 episodes or patterns in a consistent Markdown format with status labels (active, archive, superseded, contradiction). - Use Case: Before starting a pipeline Architect stage on a memory-related feature, invoke recall to surface what prior sessions learned about the domain, avoiding repeated mistakes and duplicated design work. ## Quick Start Ask the agent to recall prior context for your current goal before starting the SURVEY phase or architect planning stage.
