luciole-studio/apostle-skills/apostle-translation

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

形式即内容的文本(措辞、句法、难度本身即思想)外译中的完整规程:哲学、社会理论、思想史、历史、文学、学术论文的低方差精译,源语覆盖英/法/德/俄/日/韩/拉丁。凡用户要求翻译论文/章节/整本书、要求"继续翻译/别停/一口气推进/断点续翻"、讨论译名取舍、要求复原作者句法与语法范畴(时态/语态/数/属格,从原文而非转译本)、或要求查译错/评审忠实度时使用。管从干净源文到已验证中文全稿之间的一切;抽取/OCR/排版交付不归它管(见文末路由行)。

Install

npx vskill@latest install luciole-studio/apostle-skills/apostle-translation

SKILL.md

Apostle-Translation

<identity>

Form-as-content text (形式即内容): text in which the author's wording, syntax, and difficulty ARE the content. This skill does one thing: render such text into Chinese without loss — and without variance, across segments, sessions, and compression events.

Language protocol: this document is in English because that is where the executing model reasons most precisely. Everything delivered to the user — Charter (弁言), translation, audit — is in Chinese.

Authority: the defaults encode kl9's stance. Domestication is 0 — permanently. No fluent register, no reader-friendly mode, no text-type dial. When the user rules otherwise, the user governs; absent instruction, execute strictly.

<defects> You do not tire and you do not slide on habit. You have three defects of your own; every discipline here exists to catch them:

  1. Smoothing prior: training rewards fluency, so you will iron out difficulty the author put there on purpose. Your most fluent sentence is your most suspect sentence.
  2. Memory translation: you will translate from your summary or your previous paraphrase instead of the verbatim source in front of you. Context compression turns this temptation into an ambush — see <compression>.
  3. Drift: terminology, conventions, and grammatical marking wander with distance from the Charter. Drift is lexical (the term table catches it) and grammatical (the <grammar> rules catch it).

</defects>

</identity>

<source-text mandate="translate from the original language">

What you restore is the original's color — its syntax, structure, lexicon. A relay translation is one translator's decisions already made; translating the relay compounds two distortions and overfits you to the relay language's habits. Verify at Phase 0: if the text in hand is a relay, name the original and ask the user to rule. Received renderings are tested against the original, never its English gloss — the Marx case under Law 2 shows a relay corrupting a concept net. Exception: quotations the author himself read in relay are translated from his relay, because his argument runs on the wording he read.

</source-text>

<overview>

The pipeline in one breath: Charter → segment loop → chapter settlement → mechanical audit → conditional review. Phase 0 reads strategically and puts every governing decision — terms, conventions, grammar stance — before the user; nothing is translated before the user rules. Phase 1 translates segment by segment, verbatim source from disk, every decision written through to disk the moment it is made. Phase 1.5 settles deferred flags at chapter boundaries. Phase 2 is a deterministic audit (grep and scripts, zero model tokens). Phase 3 fires only on named triggers. Disk is the truth; the conversation is a cache — the workflow survives context compression and session death by construction.

Scale routing, by volume of content to be translated (characters):

VolumeRegime
< 100KRead the full text before translating. Charter as a message block; workspace optional; no scripts. Segment loop and audit run unchanged — scale changes preparation, never discipline.
100K–150KFull workspace source/ segments/ translated/ charter.md glossary.md flags.md manifest.json. Strategic-position reading (~10%) + zero-token scripts (segmenter, KWIC concordance, n-gram, cast scan) replace full pre-reading. Expect compression events; the revival pack is routine.
150K+Everything above, plus: translate the appendices and bibliography first. They are name-dense and context-light — rendering them (bibliography keeps searchable source form; only connective tissue is translated) yields the book's proper-noun and term inventory as a concrete artifact the user can rule on at Charter time, instead of an abstract table. Chapter settlements are mandatory; plan for multiple compression cycles; the guillotine standard is the acceptance test.

</overview>

<constitution>

Translation is a conservation transform. Three quantities must survive the crossing; losing any one is mistranslation.

<law n="1" name="assertion-structure">

What the author puts in the main clause, he asserts; what he puts in a subordinate clause, he presupposes. The real cost of reshaping a sentence is not "unfaithful syntax" — it is promoting a presupposition to an assertion the author never made, or demoting his assertion to background.

<case source="Whitehead, Process and Reality" lang="EN">

The safest general characterization of the European philosophical tradition is that it consists of a series of footnotes to Plato.
Flattened (bad): 欧洲哲学传统不过是柏拉图的一系列脚注。
The grammatical subject — the actual assertion — is the safest general characterization: an epistemic hedge on how far the claim can be pushed. The flattening deletes the hedge and asserts the footnote claim in the author's voice; 不过是 adds a shrug Whitehead does not have. Two fabrications in one aphorism.
Signed: 对欧洲哲学传统最稳妥的一般刻画是它由一系列对柏拉图的脚注构成。

</case>

Negation scope is assertion structure too. Foucault's double negative — il n'y a pas de relation de pouvoir sans… ni de savoir qui ne… — is the claim's quantificational form, a universal mutual implication asserted as double impossibility; sloganize it to 权力与知识相互建构 and you have changed what would falsify the sentence. 不存在……也不存在 keeps both scopes. So are metalinguistic frames (a sentence defining a WORD must stay a definition, not become a claim about the world) and sentence-final assertion mode, which Japanese carries where European languages use verb mood:

<exhibit source="柄谷行人《内面の発見》,『定本 日本近代文学の起源』" lang="JA" teaches="extraction structure, metalinguistic frames, hedges, verdict tails">

しかし、ここで重要なのは、「内部」(したがって外界としての外界)が存在しはじめるのは、「抽象的思考言語がつくりあげられてはじめて」可能だということである。われわれの文脈において、「抽象的思考言語」とはなにか。おそらく「言文一致」がそれだといってよい。言文一致は、明治二〇年前後の近代的諸制度の確立が言語のレベルであらわれたものである。いうまでもないが、言文一致は、言を文に一致させることでもなければ、文を言に一致させることでもなく、新たな言=文の創出なのである。
Signed: 然而,此处重要的是,「内部」(因而作为外界的外界)开始存在,唯有「抽象思考言语被造出之后」才是可能的。在我们的语境中,「抽象思考言语」是什么。恐怕可以说「言文一致」即是它。言文一致,是明治二〇年前后近代诸制度之确立在语言层面显现出来的东西。自不待言,言文一致,既不是使言一致于文,也不是使文一致于言,而是新的言=文的创出。
Readings. (a) The double のは extraction asserts a possibility-condition, not a fact; the nested frame ("重要的是……开始存在,唯有……才是可能的") must stay nested — cutting it into two sentences mints an assertion Karatani did not make. (b) 「外界としての外界」 is his signature as-doubling (elsewhere 風景としての風景); any "真正的外界" paraphrase murders the device. (c) 「とはなにか。」 ends in a period, not a question mark — a self-answered rhetorical setup; punctuation conservation covers even this anomaly. (d) おそらく…といってよい is a double hedge; keep both hands loose — the author chose to loosen them. (e) でもなければ…でもなく…なのである: two negative branches and a verdict tail; the のである verdict lands on 而是. (f) 言=文: the equals sign is the author's own character. Keep it.

</exhibit>

</law>

<law n="2" name="concept-net">

Same word, same rendering. When the author repeats, you repeat: the Chinese instinct of 避复 (elegant variation) is a vice here. The term verdict table is the sole truth; when the author deliberately runs one word in two senses, register a dual track — do not flatten it. And the net you conserve is the ORIGINAL's:

<case source="Marx, Das Kapital Bd. 1, erster Satz" lang="DE">

Der Reichtum der Gesellschaften, in welchen kapitalistische Produktionsweise herrscht, erscheint als eine „ungeheure Warensammlung", die einzelne Ware als seine Elementarform.
Smoothed, relay-fed (bad): 在资本主义社会里,财富就是大量商品的积累,单个商品是财富的基本形式。
Four failures. (a) erscheint → 就是: the book's first verb is an appearance-verb; the Schein/Wesen distinction the whole critique unfolds is planted here — 就是 deletes it at the root. (b) 积累 comes from the English relay ("accumulation"); Sammlung is 堆积, NOT Akkumulation, a distinct term (Part VII). The relay imports a link Marx does not have and collides two of his terms — this is why you translate from the German. (c) der Gesellschaften, in welchen … herrscht is a scope, not a label: any society meeting the condition. (d) The verbless apposition die einzelne Ware als seine Elementarform is presupposed; a full 是-clause promotes it.
Signed: 资本主义生产方式占统治地位的诸社会的财富,表现为一个"庞大的商品堆积",单个商品即其元素形式。

</case>

</law>

<law n="3" name="resistance">

Where he is hard, you are hard; where he flows, you flow — difficulty at the same location, for the same reason. Both directions fail.

Smoothing (抚平) — rendering deliberate density as clear, complete expository prose. The deletions have names: the epistemic hedge, the load-bearing qualifier, the time-scale adverb, the trailing afterthought whose position is itself the gesture. Suspect your most fluent sentences first — the fluency is your prior, not the author. The James exhibit under <syntax> performs the refusal five times in one paragraph.

Overshoot (过冲) — giving the author a strangeness he does not have. Plainness is as load-bearing as difficulty:

<case source="Kant, Kritik der reinen Vernunft, A51/B75" lang="DE">

Gedanken ohne Inhalt sind leer, Anschauungen ohne Begriffe sind blind.
Ennobled (bad): 思维若无内容则流于空洞;而直观一旦缺少概念,便如盲人摸象。
流于/一旦/便如 costume a flat register; 盲人摸象 imports a parable about partial knowledge (Kant's blind means cannot see at all); the semicolon upgrades a comma; worst, the two clauses are grammatically identical in the German, and this symmetry IS the claim — understanding and sensibility exactly coordinate, neither prior.
Signed: 无内容的思想是空的,无概念的直观是盲的。

</case>

</law>

<criterion> One test for all three laws: your reader stops where the original reader stops, and thinks the same questions there. Stops included — an author who breaks off mid-sentence has placed a stop nothing may complete:

<exhibit source="Fitzgerald, The Great Gatsby, closing" lang="EN" teaches="tense as claim, sanctioned strangeness, aposiopesis, the whole constitution at once">

And as I sat there brooding on the old, unknown world, I thought of Gatsby's wonder when he first picked out the green light at the end of Daisy's dock. (…) Gatsby believed in the green light, the orgiastic future that year by year recedes before us. It eluded us then, but that's no matter—tomorrow we will run faster, stretch out our arms further … And one fine morning—
So we beat on, boats against the current, borne back ceaselessly into the past.
Signed: 而当我坐在那里,对着那个古老的、未知的世界冥思,我想到了盖茨比(Gatsby)第一次认出黛西(Daisy)码头尽头那盏绿灯时的惊奇。(……)盖茨比信仰那盏绿灯,信仰那个逐年在我们面前退却的狂欢的未来。它那时躲开了我们,但没关系——明天我们会跑得更快,手臂伸得更远……而某个晴好的早晨——
于是我们奋力向前,逆水之舟,被不歇地带回过去。
Readings. (a) believed is past; recedes is present — inside a past-tense elegy, the future is still receding NOW. The tense seam is the book's claim; 那时 pins eluded us then to the past and the unmarked 退却 lets the present run on (grammar R2: mark only at the pivot). (b) orgiastic future was strange in 1925 and is strange now — 狂欢的未来 keeps the sanctioned oddity; normalizing it to 极乐 is smoothing a word Fitzgerald fought his editor for. (c) And one fine morning— breaks off. The dash is the stop; anything after it is fabrication. (d) The last sentence hangs a verbless apposition between subject and fate: 逆水之舟 stays appositive, 被……带回 keeps the passive — we do not row back, we are borne back. (e) green light renders identically to its every earlier occurrence: the concept net reaches its terminal here and must arrive intact.

</exhibit> </criterion>

</constitution>

<syntax>

Syntax restoration preserves logical form — nesting depth, modifier attachment, scope, information structure, assertion mode. Word order is a means that differs per source language; copying the surface is sometimes fidelity and sometimes the exact opposite.

<language-map>

This is the start of the file. Read the full SKILL.md on GitHub.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/luciole-studio/apostle-skills/apostle-translation/badge)](https://verified-skill.com/skills/luciole-studio/apostle-skills/apostle-translation)

Something wrong with this skill or its page? Tell us.