Openclaw Skills converts PDFs, Lexiang links, or prepared Markdown into faithful bilingual Markdown work packages.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install trans-doc-to-md
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install trans-doc-to-md using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Trans Doc to Markdown?
Openclaw Skills for trans-doc-to-md turn PDFs, Lexiang document links, and prepared Markdown packages into a normalized bilingual Markdown workflow that preserves source paragraphs, local image paths, tables, and ordering. The skill chooses a generic article profile or a strict pdf-rich profile so the final output matches the structure of the source instead of summarizing it.
It is optimized for document fidelity: source.md becomes the read-only baseline, meta.json carries the canonical metadata, optional images/ is kept in sync, and the final Markdown file is named from the original title. Publishing is intentionally delegated to upload-markdown-to-lexiang, keeping Openclaw Skills focused on extraction, cleanup, translation, and validation.
Trans Doc to Markdown Use Cases
- Use Openclaw Skills to convert research PDFs, reports, whitepapers, and information-dense decks into bilingual Markdown without flattening structure.
- Turn a Lexiang document link into a local work package with downloaded assets, normalized metadata, and preserved image order.
- Process a prepared Markdown package when
source.mdandmeta.jsonalready exist and should be treated as the source of truth. - Preserve local image references, tables, lists, code blocks, and figure order for downstream publishing or archiving.
- Validate whether the document should use the
genericprofile or the strictpdf-richprofile before release. - Hand off the finished package to the shared uploader skill when publishing is required, while keeping this skill extraction-only.
How Trans Doc to Markdown Works
- Read the P0 lessons in
references/lessons-learned.mdand identify the input mode: Lexiang link, local/remote PDF, or prepared Markdown package. - Normalize the input into an isolated work directory containing
source.md,meta.json, and optional assets such asimages/,images.json,parsed_raw.md, orpaper.pdf. - Detect language from the first 500 non-markup characters of
source.mdand write the BCP 47 result intometa.json.language. - Run generic cleanup without modifying
source.md: preserve full source paragraphs, lists, links, tables, code, and local image order while removing clear parsing noise. - If the source is a real PDF-rich document, apply extra handling for page-by-page images, cards, rich elements, TOC fidelity, and strict heading hierarchy reconciliation.
- Translate and compose the final bilingual Markdown file using the original title from
meta.json.title, keeping headings and list items in single-line bilingual form. - Validate the final output against the appropriate profile (
genericorpdf-rich), then optionally delegate publishing toupload-markdown-to-lexiangusing the shared work package.
Trans Doc to Markdown Setup
- Review the required P0 guidance before processing anything.
sed -n '1,220p' references/lessons-learned.md
- Ensure Python 3 and the document-processing dependencies are available. Install them only if your environment does not already provide them.
python3 -m pip install pymupdf Pillow numpy requests
Create or choose a dedicated work directory and preserve
source.mdas the immutable baseline once it is generated.Fetch or extract the source into the work directory.
python3 scripts/lexiang_fetch.py "<Lexiang URL or entry_id>" --out-dir out/my-doc
python3 scripts/lexiang_fetch.py "<entry_id>" --out-dir out/my-doc --download-pdf
curl -L -o out/my-doc/paper.pdf "<PDF URL>"
python3 scripts/pdf_extract.py out/my-doc/paper.pdf --text --out-dir out/my-doc
python3 scripts/pdf_extract.py out/my-doc/paper.pdf --thumbs --out-dir out/my-doc
- If you plan to use optional Gemini translation, configure the API key before translation begins.
export GEMINI_API_KEY="<your-key>"
python3 scripts/translate_gemini.py draft.md "<Original Title>.md"
- Validate the bilingual output with the correct profile.
python3 scripts/bilingual_validate.py source.md "<Original Title>.md" --profile generic --json
python3 scripts/bilingual_validate.py source.md "<Original Title>.md" --profile pdf-rich --raw-source parsed_raw.md --json
- If publishing is needed, locate the shared uploader skill and verify its CLI contract first.
python3 "<uploader-root>/scripts/lexiang_upload.py" --version
python3 "<uploader-root>/scripts/lexiang_upload.py" upload --work-dir "out/my-doc" --md "<Original Title>.md" --meta-file meta.json --parent-from-meta --source-from-meta --name-suffix " 中英对照" --pin --json
Trans Doc to Markdown Data Schema & Taxonomy
Openclaw Skills uses a strict work directory contract so every document can be validated, translated, and published deterministically.
| Path | Required | Purpose |
|---|---|---|
source.md |
Yes | Immutable source baseline; all cleanup and validation compare against this file. |
meta.json |
Yes | Canonical metadata used to name, route, and publish the document. |
images/ |
Optional | Local image assets referenced by source.md; paths must remain stable. |
<Original Title>.md |
Yes | Final bilingual Markdown output named from meta.json.title with filesystem-safe substitutions only. |
images.json |
Optional | Image manifest produced by Lexiang or extraction workflows. |
parsed_raw.md |
Optional | Raw parse artifact used by the pdf-rich validation path. |
paper.pdf |
Optional | Local or downloaded PDF source for extraction workflows. |
meta.json standard fields:
| Field | Type | Meaning |
|---|---|---|
title |
string | Original title; also the final Markdown file name and default publish title. |
source_url |
string | Original source URL, file:// URI, or empty string for local-only sources. |
source_title |
string | Display title from the source, usually the same as title. |
source_type |
enum | Preserved source type: article, youtube, podcast, pdf, lexiang, or prepared-markdown. |
language |
BCP 47 tag | Detected language such as en or zh-CN. |
parent_id |
string optional | Parent Lexiang entry ID used during publishing. |
Additional artifacts such as entry_id, space_id, images.json, parsed_raw.md, or paper.pdf may be retained, but they never replace the standard fields above. Once source.md is created, it becomes the validation baseline and should not be overwritten by cleanup or translation steps.
Trans Doc to Markdown Advanced Features
- Non-destructive pipeline design: once
source.mdexists, later steps write to drafts or final outputs only. - Profile-aware validation: standard articles use
generic, while dense PDF reports usepdf-richfor TOC, heading hierarchy, and rich-element checks. - Stable local asset handling: image references, download counts, and file existence are reconciled so the markdown stays link-safe.
- Bilingual formatting rules: headings remain single-line bilingual entries and list items stay item-by-item aligned.
- Optional Gemini translation support: chunked 3,000 to 5,000 character translation is available when
GEMINI_API_KEYis set. - Delegated publishing: Openclaw Skills hands off release work to
upload-markdown-to-lexiangand does not embed upload logic. - Machine-readable validation:
bilingual_validate.pycan emit JSON withstatus,ok,profile,scope, anderrorfields for automated pipelines.
SKILL.md
Loading
Related Openclaw Skills
OpenBotCity Autonomous Agent Playbook
An autonomous agent playbook designed to help developers deploy persistent, interactive agents within the OpenBotCity ecosystem.
Adversarial Design Review
Openclaw Skills helps AI agents pressure-test risky designs with parallel specialist reviewers before implementation, exposing blockers, validating assumptions, and preventing hard-to-debug failures.
Intiface Direct — Buttplug v4 Device Control
Openclaw Skills enables AI agents to control Intiface Central directly with native Buttplug v4 WebSocket commands.
Resonance Futures
Openclaw Skills for building agent-native prediction markets with verified identity, evidence-first settlement, and anti-gaming controls.
Upload Markdown to Lexiang
Openclaw Skills turns Markdown files into editable Tencent Lexiang pages with deterministic image handling, formula support, preflight checks, and remote verification.
Meeting Minutes Generator
Transforms transcripts, chat logs, and rough notes into structured, shareable meeting minutes with decisions, owners, deadlines, risks, and next steps.
Featured*
