What problem does it solve? Long AI coding sessions lose context to compaction, wiping out decisions made hours earlier. This Skill writes a mid-session checkpoint into the session's registered notes.md so decisions, files touched, and next steps survive compaction while the session stays open. ## Core Features & Use Cases - Session checkpointing: Resolves the active session ID, finds its registered notes.md (with a relink fallback for resumed sessions), and updates Decisions, Files touched, Open questions, and Next steps sections. - Frontmatter refresh: Auto-attaches PR links (via gh CLI and conversation scan) and ticket IDs, recording ticket statuses, without overwriting existing values. - Knowledge distillation and skill proposals: Optionally promotes durable decisions to a knowledge vault and proposes reusable skills, both gated behind configuration flags. - Use Case: Midway through a long refactoring session, run the checkpoint before context compaction so the next conversation resumes with full decisions and a prioritized next-steps queue. ## Quick Start Ask the AI to save the current session checkpoint so the notes file captures decisions and next steps before context compaction.
