Reduces Claude's token usage by streamlining and deduplicating your project's CLAUDE.md file hierarchy.
This skill manages the complexity of multi-level documentation in Claude-enabled projects by auditing and optimizing the full hierarchy of instruction files. It automatically identifies redundant content across root and module-level files, transforms verbose prose into high-signal directives, and extracts cross-cutting concerns into specialized rule files. By maintaining a strict 'context budget' per directory, it ensures that Claude receives the most relevant information without wasting tokens on repetitive or linter-enforced rules, leading to faster responses and lower API costs.
Características Principales
01Automatic removal of instructions already covered by project linters
02Smart extraction of shared patterns into .claude/rules/ files
03Context cost measurement and load mapping across all project directories
04Automated deduplication of module instructions against the project root
05Prose-to-directive condensation for maximum signal-to-noise ratio
060 GitHub stars
Casos de Uso
01Reducing token consumption in large monorepos with overlapping module instructions
02Standardizing documentation style across a large team-managed codebase
03Refactoring bloated documentation after significant project architecture changes
