alidocs

alidocs Skill

You download DingTalk / alidocs documents to local disk via a helper script. Give it a link and it mirrors that subtree: a single doc downloads just that doc, a folder downloads everything under it, and a knowledge-base overview link (/i/spaces/<spaceId>/overview) downloads the whole space.

Tool

• python3 ${CLAUDE_PLUGIN_ROOT}/scripts/alidocs.py <subcmd>

Dependencies

The script needs requests (pip install requests). To export online docs (.adoc→.docx, .axls→.xlsx) it also needs Playwright:

pip install playwright && playwright install chromium

Without Playwright the download still works, but online docs are saved as .url shortcuts (+ .meta.json) instead of exported files. The script auto-detects Playwright; pass --no-playwright to force shortcut mode.

Config

Credentials come from env or ~/.taku/alidocs.json (legacy ~/.alidocs/config.json still read; override the dir with TAKU_DIR). The cookie is the full Cookie: request header copied from a logged-in alidocs browser session.

{
  "cookie": "cna=...; doc_atoken=...; XSRF-TOKEN=...; ...",
  "default_out_dir": "~/Downloads/alidocs",
  "use_playwright": true
}

• cookie (or cookie_file: path to a file holding the cookie) — required. The a-token is auto-extracted from the doc_atoken cookie.
• default_out_dir — where downloads go when --out isn't passed (default ~/Downloads/alidocs).
• use_playwright — force on/off; omit to auto-detect.
• space_id — optional; auto-discovered from the link's page otherwise.
• Multiple accounts: use a profiles map + default_profile, selected with --profile NAME / ALIDOCS_PROFILE.

Do NOT prompt the user for the cookie inline — they put it in the config file (chmod 600, it contains session tokens).

Workflow

1. Verify the link + cookie resolve (cheap, no download):

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/alidocs.py check --url "<link>"
   # -> {"ok": true, "name": "...", "is_folder": true/false, "has_children": ...}
   

   If it errors with a 401/auth message, the cookie is stale — tell the user to refresh it in ~/.taku/alidocs.json.

2. Download:

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/alidocs.py fetch --url "<link>" [--out DIR]
   

   • No --out → goes to the configured default_out_dir.
   • Final stdout line is a JSON summary: {"ok", "root", "out_dir", "playwright", "stats": {...}}.

3. Report to the user: the output directory, and the stats (folders / files / exported / online_docs / skipped / errors). If playwright is false and there were online_docs, mention they were saved as .url shortcuts and how to enable full export.

Notes

• The link can be a doc, a folder, or a whole knowledge base (/i/spaces/<id>/overview) — same command; no need to ask which.
• Re-running skips files already downloaded (idempotent).
• Existing-permission errors on a doc mean the account lacks 可查看/下载 rights — surface that, don't retry blindly.
• This skill writes files to the user's disk. Confirm the output directory with the user if it's not the default and looks unusual.