What problem does it solve? Squad agents load their full context history on every spawn, producing 34–74KB payloads where 82–96% is stale noise. This Skill defines a three-tier memory model that cuts context size by 20–55% per spawn by loading only what the current task needs. ## Core Features & Use Cases - Three-Tier Memory Model: Hot tier (2–4KB current session context, always loaded), Cold tier (8–12KB summarized cross-session history, loaded on demand), and Wiki tier (durable structured knowledge like ADRs and conventions, loaded selectively). - Spawn Template Pattern: Provides a ready-to-use spawn prompt template that defaults to Hot-only loading, with --include-cold and --include-wiki flags for tasks needing history or domain knowledge. - Scribe Integration: Defines how the Scribe agent automates tier promotion, compressing Hot to Cold at session end and promoting Cold to Wiki after a 30-day rolling window. - Use Case: When resuming interrupted work, spawn an agent with Hot context plus the Cold summary file instead of replaying full transcripts, cutting token usage while preserving relevant decisions. ## Quick Start Spawn an agent using the tiered-memory template with only the Hot context section filled in, and add the Cold or Wiki sections only when the task requires prior history or domain reference docs.
