SureForge
SkillAn instruction-only skill that makes agents research before they ask, plan before they build, and verify before they call anything done.
- workflow
- verification
- quality
- skill
A Chinese prose-quality skill that makes AI write modern Chinese worth reading — explicitly not an AI-flavor remover.
Cídá (辞达, "say it well" — from Su Shi's letter on prose) is a Claude Code skill for modern Chinese writing quality. Its stance is unusual: it refuses the humanizer genre outright — no gaming AI detectors, no fake imperfections, no imitating living authors, no banned-word lists, no percentage scores — and instead optimizes readability, conversational force, rhythm and thought density through eleven principles (quality before humanization, audience before style, facts before flourish, mechanism before rule) and a default register of "plain Chinese that gets the point across," where any more literary treatment needs a contextual reason. The skill routes through a SKILL.md entry into eight workflows (topic to article, full rewrite, paragraph polish, oral transcript to article, deep rewrite with P0-to-P4 diagnosis, text diagnosis, platform adaptation, style calibration), backed by a knowledge base of discourse mechanisms and anti-patterns across nine areas, seven platform adapters and a ten-dimension style space. Version 1.0.0, MIT, by a Chinese content creator; same author as the in-library gongwen-gbt9704-skill.
Chinese-language writers and teams who want AI-assisted drafting that reads naturally without chasing humanizer tricks — meeting transcripts into publishable articles, drafts tightened with stated reasons, or a personal style profile learned from your own archive.
git clone https://github.com/mizzlelover/cida.git bash cida/install.sh # copies the skill to ~/.claude/skills/cida # then just ask: # 帮我把这篇文章改得更自然 # 把这段会议转录整理成可发表的文章 # Codex / OpenCode: clone into the project root; AGENTS.md is read automatically
Built for modern Chinese writing — the knowledge base, register rules and platform adapters are Chinese-first, so it is the wrong tool for English prose. Being instruction-only, output quality depends on how well your model follows the SKILL.md routing; the Calibrate path extracts style parameters from your own texts, so feed it real writing rather than samples. v1.0.0 with the structure still settling — re-read the changelog before upgrading. MIT.
Repository
mizzlelover/cidaCharted
An instruction-only skill that makes agents research before they ask, plan before they build, and verify before they call anything done.
Guizang's agent skill turns your repo's recent updates into a finished promo video — real components, code-rendered animation, an original code-written soundtrack — no video editor involved.
An agent skill that geolocates a photo with nothing readable in it — OSM geometry, elevation skylines and satellite tiles turn “where was this taken?” into a measured answer with an error radius.