genshijin
Compress Japanese responses to cut token usage 75%
All Skills in This Repository (7)
Pure Emerald Level Indicatorsgenshijin-compress
Compress Markdown memory files into genshijin format with backups and validation.
genshijin-crew
Select compressed subagents for code exploration, edits, and diff reviews.
genshijin-commit
Generate Conventional Commits-compliant commit messages with strict formatting constraints.
genshijin
Compress Japanese AI replies into 丁寧, 通常, or 極限 modes for token efficiency.
genshijin-review
Compresses pull request feedback into line-referenced comments with severity tags.
genshijin-help
Display genshijin mode triggers and sub-skill commands in one-shot help card.
genshijin-stats
Report real token usage and estimated savings from Claude Code session logs.
Frequently Asked Questions
FAQPage SchemaHow to install genshijin?▼
Run `npx skills add InterfaceX-co-jp/genshijin --all -g -y` in your terminal to install all skills globally.
How to reduce token usage in Japanese AI chats?▼
Activate genshijin mode with /genshijin to strip honorifics and filler from responses, cutting token usage by roughly 75% while keeping technical content exact.
What compression levels does genshijin offer?▼
Three levels: polite (keeps honorifics, removes filler), normal (default, keyword-style fragments), and extreme (maximum compression with abbreviations and arrows).
Does genshijin work with Cursor and Copilot?▼
Yes. It ships ready-made rule files for Cursor, Windsurf, Cline, GitHub Copilot, and any agent reading AGENTS.md.
Can genshijin compress CLAUDE.md memory files?▼
Yes. The genshijin-compress skill rewrites memory files in compressed form, backs up the original, and refuses to touch secrets like .env or SSH keys.
Related Repositories in Software Engineering
View All in Software Engineering→openclaw
Run a personal AI assistant across your devices and chat apps
superpowers
Gives coding agents a disciplined workflow from idea to merged code
react
AI agent skills for building, testing, and porting React core