What problem does it solve? AI coding sessions lose all context when they end, forcing developers to re-explain architectural decisions, conventions, and task progress every new session. This Skill automatically saves session state to structured .rune/ files and reloads it at the start of the next session. ## Core Features & Use Cases - Save Mode: Appends architectural decisions, coding conventions, task progress, and session summaries to .rune/decisions.md, .rune/conventions.md, .rune/progress.md, and .rune/session-log.md, then commits them to git. - Load Mode: Reads all state files at session start, verifies integrity, loads active invariants via a Node.js script, and presents a structured summary with the recommended next task. - Checkpoint Mode: Creates an exact resume point in .rune/checkpoint.md capturing git state, load-bearing decisions, and ordered remaining tasks for mid-work handoffs. - Detach Mode: Manages asynchronous oracle-model dispatches via idempotent pending records in .rune/oracle-pending/ with timeout and orphan cleanup. - Use Case: A developer ends a session mid-refactor; the next session automatically loads the checkpoint, sees the branch and uncommitted files, and resumes from the exact step where work stopped. ## Quick Start Ask the agent to save the current session's decisions and progress to the .rune directory so the next session can resume where this one left off.