What problem does it solve? Long AI-assisted coding sessions lose all progress when context is cleared, compacted, or the session ends, forcing you to re-explain everything from scratch. This Skill captures the conversation state—files modified, decisions made, next steps—into a restorable markdown checkpoint. ## Core Features & Use Cases - Context Checkpoint Creation: Generates a structured markdown marker in .agent/.context-markers/ capturing conversation summary, loaded docs, modified files, technical decisions, and next steps. - Theory of Mind Capture: Records user intent, stated preferences, corrections, and belief state so restored sessions avoid repeating mistakes. - Verification & Logging: Verifies the marker file exists and is non-empty, computes an MD5 checksum, and logs creation to .agent/.marker-log. - Use Case: Before a risky refactor, say "mark this point"—if the refactor fails, restore the marker in a new session and resume exactly where you left off without re-explaining the OAuth implementation you just finished. ## Quick Start Ask the assistant to save your progress by saying "create a checkpoint before my refactor" and it will write a restorable context marker file.
