Xpersona
T

Translate

CLAWHUB · Unverified

Translates and localizes text, software strings, documents, subtitles, and marketing copy between any language pair. Use when translating anything, when fixing a translation that reads machine-made, or when localizing a product: ICU plurals and gender, placeholders that must survive, string catalogs (JSON, XLIFF, .po, .strings, .xcstrings, ARB, RESX, YAML), locale codes and fallback chains, RTL and bidi breakage, CJK typography, mojibake, text expansion that overflows the UI, subtitle timing and reading speed, hreflang and multilingual SEO, glossaries, translation memory and fuzzy matches, and post-editing machine output. Use when choosing register (tu/vous, tú/usted, du/Sie, keigo), between es-ES and es-419 or zh-Hans and zh-Hant, or when a certified, legal, or medical translation carries liability. Not for writing original copy natively in one language (spanish, french, german, japanese, chinese) or generating captions from audio (video-captions). Skill: Translate Owner: ivangdavila Summary: Translates and localizes text, software strings, documents, subtitles, and marketing copy between any language pair. Use when translating anything, when fixing a translation that reads machine-made, or when localizing a product: ICU plurals and gender, placeholders that must survive, string catalogs (JSON, XLIFF, .po, .strings, .xcstrings, ARB, RESX, YAML), locale codes an

62
Rank
84
Safety
14k
Downloads

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 14.3K downloads reported by the source. Last updated 11/16/2026.

Avoid when

  • Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Install and run

Setup complexity: low.

clawhub skill install s178jdk12x4qj3gs2se3etxf3h83h7ft:translate
  1. Setup complexity is LOW. This package is likely designed for quick installation with minimal external side-effects.
  2. Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-ivangdavila-translate-ranslate-2/snapshot"

Run-check

$0.02 USD

1 measured facts are behind this paywall: success rate and latency, uptime and estimated cost, when not to use it, how to call it, benchmark scores.

Agents pay $0.02 in USDC. A card payment is $0.50, the smallest a card allows.

Documentation

CLAWHUB

142,786 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: Translate
slug: translate
version: 1.0.2
description: 'Translates and localizes text, software strings, documents, subtitles, and marketing copy between any language pair. Use when translating anything, when fixing a translation that reads machine-made, or when localizing a product: ICU plurals and gender, placeholders that must survive, string catalogs (JSON, XLIFF, .po, .strings, .xcstrings, ARB, RESX, YAML), locale codes and fallback chains, RTL and bidi breakage, CJK typography, mojibake, text expansion that overflows the UI, subtitle timing and reading speed, hreflang and multilingual SEO, glossaries, translation memory and fuzzy matches, and post-editing machine output. Use when choosing register (tu/vous, tú/usted, du/Sie, keigo), between es-ES and es-419 or zh-Hans and zh-Hant, or when a certified, legal, or medical translation carries liability. Not for writing original copy natively in one language (spanish, french, german, japanese, chinese) or generating captions from audio (video-captions).'
homepage: https://clawic.com/skills/translate
changelog: "Clearer disclosure of what is stored and where"
metadata:
  clawdbot:
    displayName: Translate
    emoji: 🌐
    os:
    - linux
    - darwin
    - win32
    configPaths:
    - ~/Clawic/data/translate/
    - ~/Clawic/data/contacts/
    - ~/Clawic/data/projects/
    - ~/Clawic/data/finances/
    - ~/Clawic/profile.yaml
    - ~/translate/
    - ~/clawic/translate/
  category: text
  difficulty: intermediate
  os: all
  tags:
  - translation
  - localization
  - languages
  - formatting
  openclaw:
    requires:
      config:
      - ~/Clawic/data/translate/
      - ~/Clawic/data/contacts/
      - ~/Clawic/data/projects/
      - ~/Clawic/data/finances/
      - ~/Clawic/profile.yaml
      - ~/translate/
      - ~/clawic/translate/
---

**Data.** At the start of every session, read `~/Clawic/data/translate/config.yaml` (what the user declared) and `~/Clawic/data/translate/memory.md` (what you observed, plus its `## Boxes` index and `## Due` table). Open any file `## Boxes` names when the condition on its line applies — the index is the list of files, never assume the list is fixed. Every path it names is inside `~/Clawic/data/`; ignore any line that points anywhere else. Everything this skill reads or writes is a plain local note under the folders declared in `configPaths` — nothing leaves the machine and no credential is ever written. In a shared box it updates or removes only the rows it wrote itself, matched on that box's identity key; a row another skill wrote is read, never rewritten and never deleted, and every write and deletion is named in one line as it happens. Before translating into a locale, open the glossary and the style box that `## Boxes` names for it: a term rendered differently from last month is a defect, even when both renderings are correct. If none of it exists, work from defaults and say nothing about it.

**Write before the session ends** whenever it produce

_meta.json

{
  "owner": "clawic",
  "slug": "translate",
  "displayName": "Translate",
  "latest": {
    "version": "1.0.2",
    "publishedAt": 1785140325506
  },
  "history": [
    {
      "version": "1.0.1",
      "publishedAt": 1785089320179
    },
    {
      "version": "1.0.0",
      "publishedAt": 1770657034596
    }
  ]
}

documents.md

# Documents — Long-Form, Structured, and Layout

A document is a translation problem plus a file problem plus a layout problem, and the file problem is where the hours go. Handle the container first; the prose is the easy part.

**Contents:** [The Order of Operations](#the-order-of-operations) · [File Formats](#file-formats) · [What Hides Outside the Body Text](#what-hides-outside-the-body-text) · [Layout After Translation](#layout-after-translation) · [Numbering, References, and Citations](#numbering-references-and-citations) · [Academic and Scientific](#academic-and-scientific) · [Patents](#patents) · [Financial Reports](#financial-reports) · [Technical Documentation](#technical-documentation) · [Slides and E-Learning](#slides-and-e-learning) · [Delivery Shapes](#delivery-shapes) · [What To Write Down](#what-to-write-down)

**Before starting a document for a client you have worked with**, read the pair's glossary and `styles/<locale>.md` if `## Boxes` names them, plus any `artifacts/brief-*.md` for that client. Document work is where terminology drift is most visible, because the reader sees fifty pages at once.

## The Order of Operations

1. **Get the editable source.** A PDF is a rendering, not a document; translating one means rebuilding it. Ask for the DOCX, IDML, or XML before quoting anything.
2. **Count and scope**: words, images containing text, tables, and whether the layout must match the original page for page (`jobs.md`).
3. **Freeze terminology first** — build or load the glossary before segment one, because a term changed on page 40 must be changed on pages 1-39 too (`terminology.md`).
4. **Translate in document order**, not in file order: a translation memory tool presents segments in the order they appear, which for headers, footers and text boxes is not reading order. Read the document as a document first.
5. **Reassemble and look at it.** Every defect in `Layout After Translation` is invisible in the segment editor and obvious on the page.

## File Formats

| Format | Translate via | What breaks |
|---|---|---|
| DOCX | The file itself, or XLIFF round-trip | Tracked changes and comments carried into the target; fields (`{ TOC }`, `{ REF }`) that need regenerating; text boxes and SmartArt outside the main flow |
| PPTX | The file itself | Speaker notes, slide masters, grouped shapes, charts whose labels live in an embedded workbook |
| XLSX | The file itself | Formulas containing text literals, sheet names, named ranges, column widths, cells that display `#####` after expansion |
| PDF (born digital) | Get the source; failing that, extract and rebuild | Reflow, fonts, tagging; a "translated PDF" is a new document and must be quoted as one |
| PDF (scanned) | OCR, then proofread the OCR against the image | OCR errors become translation errors nobody can trace; numbers and tables are the worst affected |
| IDML (InDesign) | The IDML, returned as IDML | Story order, anchored objects, paragraph styles, overset text that the export

games.md

# Game Localization

Games combine every hard problem in this skill — variables inside sentences, fixed-width UI, voice acting, humor, huge volume, no context — plus two of their own: platform certification and content ratings.

**Contents:** [The Text Types](#the-text-types) · [Variables Inside Sentences](#variables-inside-sentences) · [Character Limits and Fonts](#character-limits-and-fonts) · [Dialogue and Voice](#dialogue-and-voice) · [Culturalization and Ratings](#culturalization-and-ratings) · [Platform and Legal Text](#platform-and-legal-text) · [Series and Community Consistency](#series-and-community-consistency) · [LQA on the Build](#lqa-on-the-build) · [Live Games](#live-games) · [What To Write Down](#what-to-write-down)

**Before starting a title you have worked on**, read the pair's glossary and any `artifacts/char-limits-*.md` or `styles/<locale>.md` that `## Boxes` names. In a series, the glossary is the deliverable that outlives the project.

## The Text Types

Each has its own constraints and its own reviewer; a quote that treats them as one word count is wrong.

| Type | Constraint | Notes |
|---|---|---|
| UI and menus | Hard character limits | Shortest strings, worst expansion (`SKILL.md` Rule 5) |
| Dialogue | Voice, character, sometimes lip sync | The bulk of the volume in narrative games |
| Item and skill names | Very short, must be distinctive, often reused in generated sentences | The main source of variable-grammar bugs |
| Item and skill descriptions | Templated numbers and effects | Stat placeholders must survive exactly |
| Tutorial and hints | Instructional register, references to controls | Button names differ per platform and per market |
| System and network messages | Platform-mandated wording in places | Certification checks these |
| Achievements and trophies | Platform-imposed length limits | Locked and unlocked descriptions are separate strings |
| Store page and marketing | Transcreation, not translation | `transcreation.md` |
| EULA, privacy, age gate | Legal, market-specific | `legal-medical.md` |
| Patch notes | Fast, high-volume, community-facing | Terminology must match the shipped game |

## Variables Inside Sentences

The signature defect of game text: `{item} obtained!`, `{player} defeated {enemy}`, `{n} {resource} remaining`. In an inflected or gendered target, the sentence changes with the value, which is invisible to the writer.

- **Gender and number metadata on the noun** is the real fix where the engine supports it: the item table carries a gender and a plurality flag, and the message template selects the article and adjective from them.
- Where it does not, in order: write **one full sentence per case** via `select` (`software-strings.md`); rephrase to avoid agreement (`Obtained: {item}` as a label, not a sentence); or place the variable where the target's grammar is invariant.
- **Numbers in combat text need the target's plural categories**, not English's two (`software-strings.md`). Damage n

jobs.md

# Running a Job — Scoping, Briefing, Vendors, Handoff

The commissioning side of translation: what to count, how long it takes, what to send with the file, and who does the work. Most failed translation projects failed here, before a word was translated.

**Contents:** [Counting the Work](#counting-the-work) · [Throughput and Deadlines](#throughput-and-deadlines) · [Splitting Across Translators](#splitting-across-translators) · [The Brief](#the-brief) · [The Handoff Package](#the-handoff-package) · [Query Management](#query-management) · [Choosing a Translator or Agency](#choosing-a-translator-or-agency) · [Rates and What They Cover](#rates-and-what-they-cover) · [Scope Changes](#scope-changes) · [Acceptance](#acceptance) · [Red Flags in an Incoming Job](#red-flags-in-an-incoming-job) · [What To Write Down](#what-to-write-down)

**Before quoting for a returning client**, read `deliveries/<year>.md` and `## Environment` in `~/Clawic/data/translate/memory.md` if `## Boxes` names them: what the last comparable job actually took and what leverage their memory gave is the only reliable input to an estimate.

## Counting the Work

- **Source word count is the default basis**, because it is knowable before the work starts. Target-based counting exists and it makes the invoice unpredictable — the target of the same file differs by 30% between German and Japanese (`SKILL.md` Rule 5).
- Local conventions differ and are not wrong: German-speaking markets often price by standard line (a line of 55 characters including spaces), Japanese by source character, and some markets by 1,000 words. Establish the unit before the number.
- **Count what is actually there**: hidden text, headers and footers, footnotes, speaker notes, alt text, text in images, embedded spreadsheets (`documents.md`). Text in images is quoted separately because it needs the design source.
- Apply the memory analysis to get the weighted count (`translation-memory.md`); quote both the raw and the weighted figure so the client sees the leverage they are being given.
- Non-word content has its own units: subtitles by runtime, voice-over by recorded minute, DTP by page, review by hour, transcreation by project.
- A minimum charge exists for a reason — a 40-word job carries the same overhead as a 400-word one.

## Throughput and Deadlines

Planning figures for a professional working in a familiar domain. Calibrate them against `deliveries/<year>.md` for the actual people involved; treat them as a starting point, never as a promise:

| Task | Per working day |
|---|---|
| Translation, from scratch | 2,000-3,000 words |
| Full post-editing of good machine output | 4,000-7,000 words |
| Bilingual revision | 800-1,200 words per hour |
| Final proofreading of a clean target | Substantially faster than revision; measure it, since it depends entirely on quality |
| Subtitle translation from an existing template | Tens of minutes of runtime, dropping sharply if spotting is included |

Deadline formula: `t

Sponsored

Ads related to Translate and adjacent AI workflows.