Token Audit

Audit Claude Code token usage and find where tokens are leaking. Reads ~/.claude/projects/*.jsonl transcripts AND Claude desktop cowork (local-agent-mode) sessions locally, inventories settings/hooks/skills/MCPs, runs ccusage for baseline spend, then flags the highest-$-impact leaks (hook bloat, CLAUDE.md bloat, Opus on simple turns, context rot, skill description overhead, Bash anti-patterns, cache miss storms, redundant file reads) with ballpark weekly savings for each. All analysis is local — transcripts never leave the machine. Use when the user asks to audit their Claude Code usage, find where tokens are leaking, review token spend, optimize Claude Code, or understand why they're hitting weekly limits. Triggers: 'token audit', 'why am I hitting limits', 'optimize my Claude Code', 'token leak', 'audit my usage', 'where are my tokens going'.

BayramAnnakov Updated

File contents

bayramannakov/token-audit-skill commit b97e6276d4

Frequently asked questions

npx get-skill add bayramannakov/token-audit