What problem does it solve? Over time a Claude Code setup accumulates hundreds of installed skills, stale memory files, orphaned handoffs, and abandoned worktrees, with no visibility into which artifacts are actually used or healthy. This Skill performs a full-coverage audit of the ~/.claude ecosystem and turns raw session data into prioritized cleanup actions. ## Core Features & Use Cases - Parallel multi-category scan: Orchestrates 4 subagents covering 9 artifact categories (skills, memory, handoffs, ADRs, plans, reviews, findings, tasks, worktrees), parsing JSONL session logs for real skill invocation data. - Health scoring and evidence checks: Computes 0-100% health per category on 6 radar axes, enforces A/B-evidence-backed T1 tier promotion, and flags skills whose trigger surface is mismatched (high reference count but no measured pitfall reduction). - Interactive HTML report: Generates a dark-themed dashboard with radar chart, priority-coded P0/P1/P2 action cards, and a ready-to-run cleanup script. - Use Case: Ask "my ~/.claude feels bloated, audit my setup" and receive a report showing 8.9% skill utilization, 157 skills safe to uninstall, stale worktrees to remove, and a shell script that performs the cleanup. ## Quick Start Ask the assistant to run a full ecosystem audit of your Claude Code setup and generate the health report with cleanup recommendations.
