Token Audit
Audits Claude Code instructions and configuration files to identify token bloat and optimize context consumption.
Token Audit analyzes instruction and configuration files—including SKILL.md, CLAUDE.md, and custom slash command files—to eliminate unnecessary token bloat and minimize context window overhead. It evaluates files across structured single-file and cross-file criteria, identifying issues such as monolithic structures, verbose unindexed docs, redundant cross-file rules, and oversized thinking budgets. Rather than hunting for code bugs, Token Audit delivers actionable reduction strategies like splitting files into on-demand references, tightening triggers, and trimming prose without sacrificing human readability or nuance.
主要功能
01Analyzes SKILL.md, CLAUDE.md, and configuration prose for token bloat
02Syncs checklist rules with current Anthropic prompt engineering guidelines
03Recommends structural fixes including splitting, merging, and lazy-loading references
04Provides built-in guardrails against breaking exact commands or sacrificing readability
05Evaluates single-file and multi-file architectures for cross-file duplication
062 GitHub stars
使用场景
01Shrinking oversized CLAUDE.md or global context files that exceed 200 lines
02Optimizing skill libraries to prevent accidental tool activations and save context cost
03Refactoring instruction docs prior to deploying new Claude Code workflows