renwei: open-source Claude Skill rewrites Chinese text to remove (or add) AI flavor by scenario
stingning · x · 2026-10-25
yinfangchen open-sourced "renwei", a Claude Skill that tunes Chinese text both ways: removing AI flavor per scenario, or the reverse — turning plain speech into formal prose or textbook-grade AI tone.
Scenario packs cover chat replies (drop customer-service tone, match length to the message's purpose, offer tone alternatives when the relationship is unclear), Xiaohongshu notes (keep emoji and broken layout, forbid fabricated experiences in seeding posts), formal writing (email/reports/resumes/papers: remove boilerplate and Europeanized syntax without lowering register), long-form (delete grand openings and uplift endings, don't take stances for the author), and short-video scripts (breaths, 3-second hooks, number reading).
It builds on Wikipedia's "Signs of AI writing" 33 patterns with criteria re-derived from Chinese corpora, and compares prior projects humanizer, humanizer-zh, and shuorenhua. Install by dropping the folder into /.claude/skills/.
More from coding & agent
- Local 35B authored 800 math problems under cloud-agent supervision — authoring erred 5x more than cold-solving — Individual_Page9676 · 2026-10-25
- Engineers mock subagent retreat: 'we over-engineered and need to simplify' — _AustinCalvert_ · 2026-10-25
- When should regulated AI agents hand off to humans? Devs debate triggers and context passing — late_night_deploys · 2026-10-25
- The fix for tech debt turned out to be having AGI rewrite the code every time — zack_overflow · 2026-10-25
- Four storage pivots broke my AI editor; Memos 1.0 mutates context in place — The_ONE_393 · 2026-10-25
- prime-agent vs Claude Code: fast main agent orchestrating subagents beats raw speed — samsja19 · 2026-10-25