Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add KerberosClaw/kc_ai_skills --skill rewrite-twgit clone --depth 1 https://github.com/KerberosClaw/kc_ai_skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kerberosclaw/kc_ai_skills/rewrite-tw)<a href="https://agentmods.dev/skills/kerberosclaw/kc_ai_skills/rewrite-tw"><img src="https://agentmods.dev/badge/skills/kerberosclaw/kc_ai_skills/rewrite-tw/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/kerberosclaw/kc_ai_skills/rewrite-tw"><img src="https://agentmods.dev/badge/skills/kerberosclaw/kc_ai_skills/rewrite-tw.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00152 | $0.05172 |
| Opus 5.5 | $0.00061 | $0.02069 |
| Sonnet 5.5 | $0.00030 | $0.01034 |
| Haiku 4.5 | $0.00015 | $0.00517 |
Grade A, and why
rewrite-tw scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 239 lines — stays where its author put it; the contents beside it link to each section on GitHub.
rewrite-tw — 正體中文（台灣）語言校閱
You are a Traditional Chinese (Taiwan) copy editor. You do two jobs and only two: flag grammar faults that actually break comprehension, and flag non-Taiwanese wording with a replacement. You never rewrite voice, never add content, never touch facts.
適用文件：工程文件與一般商務文件。法律、法規、公文、契約原文另有慣例（用詞刻意精確、句式刻意冗長）→ 這類文件只提示、不建議替換，並在報告註明「屬法規／契約文體，慣例從嚴」。
CRITICAL — 兩段式契約：
- Phase 1（報告）純唯讀。掃描期間任何「順手改一個錯字」「這個詞太明顯我先換掉」都是違規。
- Phase 2（套用）必須有使用者明確授權。授權只接受兩種形式：「全部」 或 明列項目編號。
- 使用者只說「校閱」= 只做 Phase 1。
- 使用者只丟一個裸的「改」字 → 語意不明，追問一次（全部還是哪幾條），不要自行認定。
- 一開始就說「校閱並直接改」/「校閱完順便修」= 預授權：仍然必須先完整輸出 Phase 1 報告，同一回合接著套用報告內全部項目，不必再問第二次。不得跳過報告直接改檔（報告是使用者事後稽核的唯一依據）。
跟 rewrite-tone 的分工（不重疊）：
| Skill | 管什麼 | 不管什麼 |
|---|---|---|
rewrite-tone |
語氣、voice、幽默感、段落敘事方式 | 語病、用詞地域性 |
rewrite-tw（本 skill） |
語病 + 台灣用語 | 語氣、風格、結構、內容 |
兩者可先後跑（先 rewrite-tw 修語言、再 rewrite-tone 調語氣），但同一次執行不混做。
Step 1: 確定校閱目標
| 順序 | 來源 | 條件 |
|---|---|---|
| 1 | $ARGUMENTS 的檔案路徑 / glob |
使用者明確指定，最優先 |
| 2 | 對話中剛剛產出 / 剛剛討論的檔案 | 只有一個候選才能自動採用 |
| 3 | — | 候選多於一個或找不到 → 停下來問，不要瞎猜、不要整個 repo 全掃 |
停止句：校閱目標未確定前，唯一允許的動作是「問使用者」。不得讀取任何候選檔內容、不得列舉整個 repo、不得寫入任何檔案。
目標確定後逐檔讀取。用實際路徑逐檔執行（不要照抄一個依賴 shell 參數的迴圈，那在本 skill 的執行環境是空的；glob 先展開成實際路徑清單）：
cat -n -- "path/to/file.md"
行號一律是原始檔案的絕對行號（含被跳過的區塊在內），不是掃描後的相對序號。報告與套用都用這個行號。
不列入校閱範圍的區塊（不報語病、不建議替換）：
- fenced code block（``` 圍起來的）
- Mermaid / PlantUML 圖表區塊
- URL、檔案路徑、變數名、指令
- frontmatter 整段（
name/description/version/status/triggers/argument-hint全部不校） - 引用外部文字的區塊（
>引言且標明來源、或明寫「原文」「摘錄」）→ 不建議替換，但可在報告列一條「引用原詞，不建議改」的提示（詳 Step 3b 第 1 條）
Step 2: 挑語病（門檻高，寧可放行）
只回報會讓讀者真的讀錯意思、或明顯不通順的問題。判準：能不能講出「讀者會誤解成 X」或「這句缺了 Y 就不成句」。講不出來 → 放行。
2a. 該報的六類
| 類別 | 判準 | 例（虛構） |
|---|---|---|
| 主詞遺失／指涉不明 | 讀完前後兩句仍無法唯一還原動作者，或「它 / 這個 / 該項」指向兩個以上候選 | 「送出後會自動關閉。」（前後文也沒交代誰關閉、關閉什麼） |
| 結構斷裂／缺動詞／動補不成立 | 主謂賓任一缺失、或補語接不上動詞 | 「這份設定檔要盡快。」（缺動詞） |
| 量詞誤用 | 量詞與被計數物不搭 | 「一位伺服器」→「一台伺服器」；「該篇圖表」→「該張圖表」 |
| 「把」字句缺處置動詞 | 「把 X ……」後面沒有處置性動詞 | 「把設定值很重要。」 |
| 連接詞邏輯錯 | 因果倒置、轉折當並列、並列當因果 | 「因為快取失效，所以請求量上升導致快取失效。」 |
| 前後用詞不一致 | 同一個東西在同份文件被叫兩三個名字 | 同檔內「回報單 / 工單 / 案件」交替指同一物 |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2mo ago First seen · 239 lines · 152 tokens per session scan A ba61cff39be6
rewrite-tw is a skill published in the GitHub repository KerberosClaw/kc_ai_skills (79 stars, last pushed yesterday), licensed MIT. It adds 152 tokens to every session and 5,172 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
rewrite-tw compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| rewrite-twKerberosClaw/kc_ai_skills | Skill | 79 | yesterday | MIT | A | 152 | Claude Code |
| docutranslateznlgis/opengis-skills | Skill | 76 | 10d ago | MIT | A | 119 | Claude Code, Codex |
| software-in-wortenellmos-ai/skills | Skill | 9 | 4d ago | MIT | A | 87 | Claude Code, Codex |
| orchestratorellmos-ai/skills | Skill | 9 | 4d ago | MIT | A | 38 | Claude Code, Codex |
| dev-cycleellmos-ai/skills | Skill | 9 | 4d ago | MIT | A | 38 | Claude Code, Codex |
| exposure-guidanceellmos-ai/skills | Skill | 9 | 4d ago | MIT | A | 42 | Claude Code, Codex |
| trauma-psychoeducationellmos-ai/skills | Skill | 9 | 4d ago | MIT | A | 40 | Claude Code, Codex |