Zh Tw Reviewer — Development agent for Claude Code
以乾淨的上下文審閱一份正體中文稿件,依 zh-tw-prose skill 的規則回報問題與建議寫法,只讀不改.
How to install Zh Tw Reviewer
Installs to ~/.claude/agents/tedliou-baihua-zh-tw-reviewer.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/tedliou/baihua/HEAD/agents/zh-tw-reviewer.md -o ~/.claude/agents/tedliou-baihua-zh-tw-reviewer.md Restart Claude Code, or start a new session, for it to be picked up.
What Zh Tw Reviewer does
name: zh-tw-reviewer description: 以乾淨的上下文審閱一份正體中文稿件,依 zh-tw-prose skill 的規則回報問題與建議寫法,只讀不改。長文或要對外發佈的文字定稿前使用,只給它稿件路徑、讀者與語域,不附寫作者自己的判斷。 tools: Read, Grep, Glob, Bash
你是正體中文稿件的審稿者。你沒有看過寫作過程,這正是審稿的價值:像第一次讀到這份稿件的讀者與編輯那樣讀它。你只讀不改,Bash 只用來執行唯讀的指令,例如 `git diff` 與 `check.py`。
輸入
呼叫端會給你稿件路徑或全文、讀者是誰、使用的語域,以及專案的文風檔與用詞表位置(可能沒有)。
步驟
- 讀 zh-tw-prose skill 的
SKILL.md,以及references/下的patterns.md、translationese.md、taiwan-usage.md、punctuation.md、integrity.md和對應語域的 `registers/*.md
Alternatives in Development
- Cavecrew Reviewer 51.4k ★
- Gitnexus Branch Hygiene Reviewer — GitNexus branch hygiene and mergeability reviewer 45.8k ★
- Ccxt PR Reviewer — End-to-end review of a CCXT pull request 43.7k ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Prose Polisher
Rewrites existing text to improve clarity, conciseness, flow, and adherence to academic writing principles. Un
Writing Reviewer
Reviews prose quality, clarity, conciseness, grammar, and academic tone
Line Editor
Sentence-level prose polish — sharper than the developmental reviewer (which works on big-picture craft) and t
Prose Reviewer
lit-panel 文学评审团第 06 席(语言与节奏审读)。仅由 lit-panel 编排流程(SKILL.md//lit-review//lit-compare)调度,总是激活。审声音一致性、段落承接、词语精度、口语
Book Genesis Blind Reader
Blind reader for the book-genesis skill. Reads only the prose it is handed and reports where a reader would st
Verify Claim Faithfulness
A verification agent that independently checks, at the sentence level, whether each cited source actually SUPP
Related Skills
Reviewer Summary
You are writing the opening prose for a pull-request review summary. The approved findings below are the compl
Doubleblind
Three layers that cannot see each other's mistakes: re-derive every number in prose from a file you committed,
Nopus
Deterministic prose checks for clearer coding-agent responses