What problem does it solve? When an AI conversation approaches its token limit or undergoes compaction, critical work state can be lost. This Skill defines a structured handoff protocol that preserves goals, decisions, progress, and next steps so a fresh context can resume work without information loss. ## Core Features & Use Cases - Activation Signals: Detects low-context conditions such as upcoming compaction, truncated responses, or estimated remaining tokens below 20% of the window. - Three-Phase Protocol: Preparation (persist observations via mem_save and generate a session summary), Transfer (record SDD phase, pending decisions, and queued actions), and Reconnection (restore state via mem_context and mem_search). - Standardized Handoff Format: Provides a Markdown template covering current state, last actions, next steps, relevant files, and unresolved decisions. - Use Case: During a long Spec-Driven Development session, the model detects context running low, saves a structured handoff summary to persistent memory, and the next session resumes exactly where the previous one stopped. ## Quick Start Ask the agent to generate a context handoff summary of the current session state before the conversation runs out of tokens.
