联通云盘批量上传

SKILL开发
开发 未标注权限 许可证不明确
安装这个技能

Upload files in bulk to China Unicom Cloud Drive (联通云盘 / 沃盘 / WoPan) without WebDAV, by driving its open upload API and encrypted RPC directly. Use when the user asks to push files, a folder, or a whole local directory into 联通云盘, to resolve a 联通云盘 folder ID from a human-readable path, or when OpenList/AList is mounted on WoPan and WebDAV transfers prove too slow. Also covers harvesting an existing WoPan access token out of an OpenList/AList instance.

一句话选型结论

按资料字段整理

联通云盘批量上传是开发类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类开发 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_dd794d14/unicom-wopan-upload
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

开发
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_dd794d14仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791625263
正文体量7.8 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
写代码与改 bug

按需求补代码、定位并修掉报错。

读仓库与生成改动

让模型读懂项目结构,再给出可落地的改动。

跑测试与排查故障

按现象排查问题,补测试用例并复跑。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在开发场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
联通云盘批量上传当前 未标注权限 未标注 未标注 未标注 有
技能审查员 浏览器控制 ClawHub 市场 877.9 万 OpenClaw 有
Vercel React 最佳实践 浏览器控制 GitHub 仓库 849.1 万 Codex、Claude、Cursor 有
GitHub 代码协作 联网访问 ClawHub 市场 763.3 万 OpenClaw 有
网页设计指南 联网访问 GitHub 仓库 668.6 万 Codex、Claude、Cursor 有
本体论 只读 ClawHub 市场 663.7 万 OpenClaw 有

SKILL.md 全文

7.8 KB
模型实际读到的说明书原文,来自技能仓库

联通云盘批量上传 (WoPan bulk upload)

Upload files straight to 联通云盘 through its own API. This bypasses WebDAV,
which is typically 3-10x slower (a Cloudflare-hosted OpenList in front of
WoPan measured ~0.17 MB/s vs ~0.5 MB/s direct per connection).

When to use

  • "把这些文件传到联通云盘" / "上传到沃盘" / "batch upload to Unicom Cloud Drive"
  • A folder must land at a specific path inside 联通云盘 (e.g. /OpenList/music)
  • OpenList/AList already mounts WoPan and WebDAV transfers are too slow
  • The WoPan folder ID for a human-readable path must be resolved

Do not use this for other drives (夸克 / 百度 / 阿里) — only WoPan speaks this protocol.

Prerequisites

  1. An access token. Two ways to get one:
    - From an existing OpenList/AList instance (preferred, needs no user action):
    log into the OpenList API and read the storage config — see
    references/protocol.md → "Harvesting a token from OpenList".
    This works whenever the user already has WoPan mounted there.
    - By hand: log in at <https://pan.wo.cn/>, open devtools → Network,
    find the request containing key=PcLoginVerifyCode, and read
    accessToken / refreshToken from the response.
    Note: logging into the web client afterwards may kick the driver offline;
    the mobile app does not.
  2. pyaes (preferred) or cryptography installed — used for the AES-128-CBC layer.
    pip install pyaes. Install pypinyin too if filenames may contain CJK.

Workflow

1. Resolve the target folder ID

WoPan addresses folders by opaque IDs, not paths. Walk the tree from the
personal root (&quot;0&quot;):

from wopan import WoPan
c = WoPan(access_token, refresh_token)
folder_id, created = c.resolve_path([&quot;OpenList&quot;, &quot;music&quot;, &quot;Archive.org&quot;])
# with create=True missing folders are made; returns the id of the deepest one

resolve_path returns (None, [...]) when a segment is missing and
create=False.

2. Normalise filenames

联通云盘 handles non-ASCII poorly and pinyin/ASCII names are usually wanted.
Run to_ascii() (in scripts/upload_batch.py) first: accents are folded
(Jörg → Jorg, Iñaki → Inaki) and CJK is transliterated via pypinyin.

3. Upload

python scripts/upload_batch.py \
    --src   &quot;C:\OpenList\music&quot; \
    --target OpenList/music/Archive.org \
    --storages path/to/storages.json          # or $WOPAN_ACCESS_TOKEN

python scripts/upload_batch.py --src DIR --target A/B/C --create   # mkdir as needed
python scripts/upload_batch.py --src DIR --target A/B/C --rename   # ASCII/pinyin in place
WORKERS=1 python scripts/upload_batch.py ...                       # concurrency (capped at 2)

Credentials resolve in this order — first hit wins:

  1. --storages PATH
  2. $WOPAN_STORAGES
  3. $WOPAN_ACCESS_TOKEN (+ optional $WOPAN_REFRESH_TOKEN)
  4. ./storages.json

The script lists the destination folder first and skips names already
present
, so re-running after a partial failure is safe — a verified re-run
against an already-complete folder prints uploading 0 file(s) and RESULT: PASS.

4. Verify

The script re-lists the cloud folder at the end and compares names and
sizes, both directions
, then prints RESULT: PASS or FAIL. Treat FAIL
as failure. A non-empty extra on cloud= or SIZE MISMATCH block means the
folder drifted from the local source.

Concurrency: keep it at 1-2

The uplink to WoPan is fragile and slower than it looks. Measured on a real run
(50 lossless FLAC, ~585 MB): ~0.28-0.32 MB/s per connection, ~0.58 MB/s
aggregate at 2 workers
, whole batch ≈ 20 min, 0 failures. Cap parallelism at
2; drop to 1 if transfers start failing. Concurrency is not the main failure
mode though — timeouts are, see below. Budget roughly total_MB / 0.58 seconds
and tell the user before starting a large batch.

Reporting back

After the final verify, it is worth emitting a small manifest (cloud listing
cross-checked against the local folder: counts, per-file size match, missing,
extra) plus a human-readable table of the batch. A "51 entries, missing=none"
line is far more convincing to the user than "done".

Common pitfalls

  • Never pass a (connect, read) timeout tuple to the upload POST.
    This is the single biggest trap. In this environment requests/urllib3
    applies the connect value as the socket write timeout, so any upload
    that takes longer than that dies with
    ConnectionError: (&#x27;Connection aborted.&#x27;, TimeoutError(&#x27;The write operation timed out&#x27;))
    — and raising the read value changes nothing. The failure lands at exactly
    the connect-timeout number of seconds, which is the tell.
    Always use a scalar: timeout=3600. Verified: tuple (10,300) fails a
    10 MB part at 10.4 s, scalar 300 completes it in 20.0 s.
  • session.proxies can be silently overridden by environment proxies.
    requests merges http_proxy/https_proxy from the environment into the
    request-level proxy dict, and request-level beats session-level. If a
    sandbox injects a proxy, every API call fails with ProxyError while curl
    works fine. The client sets s.trust_env = False — keep that line.
  • The response DATA field is a quoted string in the raw wire format.
    Go sees the quotes; a Python json parse has already stripped them, so
    decide "plain JSON or encrypted blob?" by trying to parse, not by checking
    for surrounding quotes.
  • totalPart uses floor division, not ceil. totalPart = max(1, size // 8MiB)
    and the final part absorbs the whole remainder, so a part may exceed 8 MiB.
    Using ceil or a bounded final part size desynchronises the part count and the
    server rejects the upload.
  • Token expiry surfaces as RSP_CODE 9999. The client auto-refreshes once
    via the api-user AppRefreshToken method when a refresh_token is available.
  • Access tokens are secrets. Never echo them into logs or the final reply.

Bundled resources

  • scripts/wopan.py — the client: AES layer, signed RPC dispatcher, chunked
    upload, resolve_path, mkdir, move, delete.
  • scripts/upload_batch.py — batch driver: token resolution, rename to ASCII,
    skip existing, parallel upload, size-aware two-way verify.
  • references/protocol.md — full wire protocol, endpoints, constants, and the
    OpenList token-harvest recipe.

Verifying through OpenList (when the user has it mounted)

An upload confirmed by the WoPan API is not the same as the user seeing it.
If OpenList sits in front of the drive, confirm independently with a WebDAV
PROPFIND (Depth: 1) on the mount path — e.g. /dav/wopan/OpenList/music/Album
— and diff names and sizes against the local folder.

Gotcha: WebDAV hrefs come back percent-encoded, so %20 / %26 / %2C
must be unquote()d before comparing, or every file looks "missing" while the
listings actually agree. A naive check will report all names missing and all
names extra at the same time — that pairing is the tell.

常见问题

5 问
答案由本站依据条目字段整理
联通云盘批量上传怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_dd794d14/unicom-wopan-upload。安装方式为未标注。装完刷新技能列表或重开客户端。

联通云盘批量上传支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

联通云盘批量上传权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

联通云盘批量上传收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

联通云盘批量上传用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。