What problem does it solve? Long, multi-phase agent conversations accumulate noisy history—failed attempts, verbose logs, abandoned branches—that degrades reasoning quality. This Skill keeps the active conversation as a clean working set by checkpointing milestones, reviewing timeline structure, and compacting stale raw history into state summaries at continuation boundaries. ## Core Features & Use Cases - Semantic Checkpoints: Create named anchors before noisy work, risky attempts, phase changes, or task switches so you can compact back to a clean state later. - Timeline Review: Inspect the structural shape of the active conversation path to choose the right compaction anchor when multiple branches or checkpoints exist. - State-Summary Compaction: Replace raw history with a compact summary capturing task state, external side effects, validation status, and next steps—without rolling back files or external systems. - Scenario References: Six reference guides cover search/research, development/debugging, planning/execution, batch work, task switching, interleaved async work, and retry/pivot patterns. - Use Case: During a long debugging session, checkpoint before each fix attempt; once root cause is confirmed, compact the noisy investigation trail into a summary and proceed to implementation with a clean working set. ## Quick Start Ask the agent to manage context during a long debugging or research session using checkpoints and compaction at phase boundaries.
