OpenClaw skill for downloading public documentation into Markdown format.
For the user, use only:
/docs onefile <url>/docs manyfiles <url>
The downloader first tries r.jina.ai and automatically falls back to direct fetch if needed.
npm installnode scripts/getdocs.js onefile <url>node scripts/getdocs.js manyfiles <url>/docs onefile <url>/docs manyfiles <url>
onefilemode: Returns a singlefull-docs.mdfilemanyfilesmode: Returns a folder with individual Markdown files plusFILELIST.md
If QMD is available, getdocs automatically creates a QMD collection after a successful download.
This lets an AI agent query large downloaded documentation through search first, instead of reading every Markdown file into context.
Typical flow:
- Download docs with
onefileormanyfiles. getdocscreates a QMD collection for the generated Markdown output.- Follow-up analysis starts with QMD search against that collection.
- The agent reads only the relevant files or line ranges returned by QMD.
The download still succeeds if QMD is not installed or indexing fails.
Detection order:
GETDOCS_QMD_BIN, if set.- A local workspace wrapper at
scripts/qmd-local.sh, if present. qmdfromPATH, for normal CLI installations.
You can disable QMD indexing with:
GETDOCS_QMD=off node scripts/getdocs.js manyfiles <url>When QMD indexing is attempted, the wrapper may include:
RESULT_QMD_STATUS=created|disabled|failedRESULT_QMD_COLLECTION=<collection-name>RESULT_QMD_COMMAND_HINT=<qmd command to search this collection>
For follow-up questions, use the command from RESULT_QMD_COMMAND_HINT and filter by RESULT_QMD_COLLECTION.
Example:
qmd search -c <RESULT_QMD_COLLECTION> "how OAuth works" -n 8 --line-numbersThen read only the selected hits, not the whole downloaded docs corpus.
- cheerio
- turndown
- turndown-plugin-gfm
MIT
Created by Denis Skripnik — blind developer working with AI agents, accessibility, automation, and Web3 tooling.
- X: https://x.com/denis_skripnik
- Telegram (ru): https://t.me/blind_dev
- VK (ru): https://vk.com/blind_dev