What problem does it solve?
This Skill prunes bloated Claude Code session files by removing progress ticks, duplicate content, stale reads, oversized tool outputs, and metadata bloat to reclaim token budget and prevent Agent Teams from losing critical context.
Core Features & Use Cases
- Prescription-driven pruning: choose gentle, standard, or aggressive prescriptions to balance safety and savings.
- Safe, audited workflow: diagnose first, perform a dry-run showing both byte and token savings, request explicit user confirmation, and create automatic timestamped backups before applying changes.
- Agent Teams protection: preserves conversation DAG integrity by never modifying uuid or parentUuid and provides token-aware diagnostics to avoid catastrophic auto-compaction loss.
- Use Case: when a Claude Code session approaches the model context limit, run a dry-run to see savings then apply a standard or aggressive prescription to restore usable tokens while keeping recoverable backups.
Quick Start
Dry-run a standard prune, review the reported token savings, and confirm to apply the treatment to the current session.