Description

Use Baidu Netdisk via mcporter + stdio MCP server with hot-reload token file credentials. Triggers when you need Baidu Netdisk operations (official 2.0 tools...

README (SKILL.md)

baidudisk-mcp

Name: Baidudisk Mcp
Author: forevershu

Use this skill to run Baidu Netdisk tools through mcporter.

1) Enable server in mcporter

Run:

bash scripts/baidudisk_mcporter.sh register

This writes a baidudisk stdio server entry into config/mcporter.json with:

command: /home/linuxbrew/.linuxbrew/bin/uv
args: --directory \x3Cworkspace>/skills/baidudisk-mcp/server run netdisk.py
env: BAIDU_NETDISK_TOKEN_FILE=~/.openclaw/credentials/baidudisk.json

Check status:

bash scripts/baidudisk_mcporter.sh check
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json list baidudisk

2) Credential file and hot update

Credential file (local only):

~/.openclaw/credentials/baidudisk.json

Expected keys:

access_token (required)
default_dir (recommended)

Example (do not commit real token):

{
  "access_token": "\x3Credacted>",
  "default_dir": "/Openclaw/baidudisk"
}

The MCP server reads this file on every tool call. So token rotation only needs editing this file; no server restart is required.

3) 2.0 tools

3.1 Official-aligned tools (主推)

四个基础列表接口与 MCP tool 对应关系：

/rest/2.0/xpan/file?method=list → file_list(dir?, limit?, order?, desc?, start?)
/rest/2.0/xpan/file?method=imagelist → file_image_list(parent_path?, recursion?, page?, num?, order?, desc?, web=1)
/rest/2.0/xpan/file?method=doclist → file_doc_list(parent_path?, recursion?, page?, num?, order?, desc?)
/rest/2.0/xpan/file?method=videolist → file_video_list_api(parent_path?, recursion?, page?, num?, order?, desc?, web=1)

另保留一个历史兼容视频工具（非官方 videolist 参数模型）：

file_video_list(dir?, recursion?, start?, limit?, order?, desc?)（基于 xpanfilelistall 过滤视频）

其他工具：

category_info(category, parent_path='/', recursion=1)
category_info_multi(categories, parent_path='/', recursion=1)
image_gettags(type=1)
image_gettags_summary(type=1, top=50)
image_search(search_type, keyword, start=0, limit=100, size?)
recent_list(category=3, start=0, limit=100, sortby?, order?, stime?, etime?, resolution='off')
file_meta(fsids, dlink?, thumb?, extra?, needmedia?, path?)
make_dir(path, parent_dir?)
file_copy(src_path, dest_dir, new_name?, ondup?)
file_copy_batch(items, ondup='newcopy', async_mode=1, chunk_size=100, dry_run=false, allow_dest_prefixes=['/Openclaw'])
file_del(path, confirm)
file_move(src_path, dest_dir, new_name?, ondup?)
file_move_batch(items, ondup='fail', async_mode=1, chunk_size=100, dry_run=false, allow_dest_prefixes=['/Openclaw'])
file_rename(path, new_name)
file_rename_batch(items, async_mode=1, chunk_size=100, dry_run=false)
file_upload_stdio(local_file_path, remote_dir?, remote_name?)
file_upload_by_url(url, remote_dir?, remote_name?, timeout_s?, max_bytes?)
file_upload_by_text(text, remote_dir?, remote_name?, max_chars?, max_bytes?)
file_keyword_search(keyword, dir?, recursion?, num?, page?)
file_semantics_search(...)（当前 unsupported stub）
file_sharelink_set(...)（当前 unsupported stub）
user_info()
get_quota(checkexpire?, checkfree?)

3.2 Legacy aliases (兼容保留)

list
search
mkdir
move
rename
delete
upload
download

4) mcporter examples

# 对齐 list：/rest/2.0/xpan/file?method=list
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_list dir=/ limit=3

# 对齐 imagelist：/rest/2.0/xpan/file?method=imagelist
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_image_list parent_path='/来自：iPhone' recursion=0 page=1 num=3 order=time desc=1

# 对齐 doclist：/rest/2.0/xpan/file?method=doclist
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_doc_list parent_path='/Openclaw' recursion=1 page=1 num=3 order=time desc=1

# 对齐 videolist：/rest/2.0/xpan/file?method=videolist
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_video_list_api parent_path='/' recursion=1 page=1 num=3 order=time desc=1
# 若根目录返回 count=0，可改 parent_path='/来自：iPhone' 或 '/Openclaw'

# 官方同名：关键词搜索
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_keyword_search keyword=invoice num=20

# 分类统计（单分类：图片=3）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.category_info category=3 parent_path=/ recursion=1

# 分类统计（多分类：JSON array）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.category_info_multi categories='[1,3,4,6]' parent_path=/ recursion=1

# 图片智能标签
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.image_gettags type=1

# 图片智能标签摘要（按 count 取前 20）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.image_gettags_summary type=1 top=20

# 图片关键字检索（search_type=2 表示汉字）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.image_search \
  search_type=2 keyword=截图 limit=20 size='c256_u256,c512_u512'

# 图片列表（imagelist，默认 web=1 返回缩略图字段）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_image_list \
  parent_path='/Openclaw' recursion=1 page=1 num=10 order=time desc=1

# 按上传时间拉取最近图片
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.recent_list \
  category=3 limit=20 resolution=off

# 官方同名：创建目录
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.make_dir path=reports/2026

# 批量移动（默认 ondup=fail，不覆盖；dest 限制在 /Openclaw）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_move_batch \
  items='[{"src_path":"/Openclaw/inbox/a.txt","dest_dir":"/Openclaw/archive"},{"src_path":"/Openclaw/inbox/b.txt","dest_dir":"/Openclaw/archive"}]' \
  chunk_size=100 async_mode=1

# 批量复制（显式冲突策略 newcopy：冲突时自动副本，不覆盖原文件）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_copy_batch \
  items='[{"src_path":"/Openclaw/inbox/a.txt","dest_dir":"/Openclaw/archive"}]' \
  ondup=newcopy async_mode=1

# 批量重命名（dry_run 先演练）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_rename_batch \
  items='[{"path":"/Openclaw/archive/a.txt","new_name":"a-20260228.txt"}]' \
  dry_run=true

# 官方同名：本地文件上传
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_upload_stdio local_file_path=/tmp/demo.txt

# 官方同名：按 URL 上传（HTTP/HTTPS，带大小/超时限制）
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_upload_by_url url=https://example.com/demo.txt remote_dir=/Openclaw/baidudisk

# 官方同名：文本上传
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.file_upload_by_text text='hello world' remote_name=hello.txt

# 官方同名：配额
mcporter --config /home/ubuntu/.openclaw/workspace/config/mcporter.json call baidudisk.get_quota

5) Batch return envelope（8.2）

file_move_batch / file_copy_batch / file_rename_batch 统一返回：

status, tool, dry_run, async_mode, ondup, chunk_size
total_items, chunks
summary（ok_chunks/failed_chunks/ok_items_est/failed_items_est）
results（每个 chunk：chunk_index/items/request_preview/response(errno/taskid/info)）
errors（失败 chunk 的 chunk_index/message/items_preview）

6) Minimal self-check（不含 delete）

请不要直接用 python3（容易缺 mcp 依赖）。

cd /home/ubuntu/.openclaw/workspace/skills/baidudisk-mcp
server/.venv/bin/python server/selfcheck_batch_filemanager.py
server/.venv/bin/python server/selfcheck_misc_readonly.py
# 下两项为真实只读联网检查（需要 token）
BAIDU_NETDISK_TOKEN_FILE=~/.openclaw/credentials/baidudisk.json server/.venv/bin/python server/selfcheck_image_recent_readonly.py
BAIDU_NETDISK_TOKEN_FILE=~/.openclaw/credentials/baidudisk.json server/.venv/bin/python server/selfcheck_lists_readonly.py

# 等价写法（如果你用 uv）
# uv run --directory server python selfcheck_batch_filemanager.py
# uv run --directory server python selfcheck_misc_readonly.py

覆盖点：

file_move_batch（10 条 + 分片降级）
file_copy_batch（ondup=fail 与 ondup=newcopy）
file_rename_batch
category_info(category=3, parent_path='/', recursion=1)
image_gettags(type=1)
四个列表基础能力（只读）：
- file_list(dir='/', limit=3)
- file_image_list(parent_path='/来自：iPhone', recursion=0, page=1, num=3, order='time', desc=1)
- file_doc_list(parent_path='/Openclaw', recursion=1, page=1, num=3, order='time', desc=1)
- file_video_list_api(parent_path='/', recursion=1, page=1, num=3, order='time', desc=1)（根目录为空时回退 /来自：iPhone / /Openclaw）

7) Safety defaults

Never store access_token in repo files.
Error messages sanitize access_token patterns.
delete/file_del rejects by default unless confirm="DELETE".
Batch move/copy restricts destination by allow_dest_prefixes（默认 ['/Openclaw']）。
file_move_batch 默认 ondup=fail（不覆盖）；file_copy_batch 默认 ondup=newcopy（冲突产出副本，不覆盖）。
file_upload_by_url only allows http(s) and rejects localhost / 内网 / 保留地址；并限制下载大小与超时。

Usage Guidance

This skill appears to implement a legitimate Baidu Netdisk MCP server, but there are several red flags you should address before installing: 1) The package metadata claims no credentials required, yet the server needs a Baidu access_token (via a token file or BAIDU_NETDISK_ACCESS_TOKEN); do not provide tokens until you verify the code. 2) SKILL.md references scripts/baidudisk_mcporter.sh but the bundle contains scripts/register_mcporter.sh which expects a non-included workspace script — registration may fail or run unexpected commands. 3) The instructions hard-code absolute paths (uv binary, /home/ubuntu workspace paths); inspect and adjust these to your environment before running. 4) The skill will write/modify mcporter config in your workspace and will read a token file from your home directory every tool call — ensure that path and the file permissions are acceptable. Recommended steps: review server/netdisk.py and the register script in full, run them in an isolated environment (or container), confirm or replace the registration commands with safe local equivalents, and only put a Baidu access_token into the token file after you are comfortable with the code. If you cannot verify the missing registration script or the hard-coded paths, treat this package as untrusted.

Capability Analysis

Type: OpenClaw Skill Name: baidudisk-mcp Version: 1.0.0 The baidudisk-mcp skill bundle provides a comprehensive MCP server for managing Baidu Netdisk operations. It implements robust security practices, including SSRF protection for URL-based uploads in `server/netdisk.py` (via `_validate_public_http_url`), automatic sanitization of access tokens in error messages (`_sanitize_message`), and a mandatory confirmation string ('DELETE') for file deletions. The code is well-structured, uses official Baidu API endpoints, and includes extensive self-check scripts to ensure functionality without evidence of malicious intent or data exfiltration.

Capability Assessment

⚠ Purpose & Capability

The code and SKILL.md implement a Baidu Netdisk MCP stdio server (netdisk.py + generated openapi client) which is coherent with the name/description. However the registry metadata claims no required environment variables or primary credential, while the runtime code and SKILL.md clearly require an access_token (via a token file or BAIDU_NETDISK_ACCESS_TOKEN). Also SKILL.md calls scripts/baidudisk_mcporter.sh but the included script is scripts/register_mcporter.sh which delegates to a non-included workspace script; the expected registration script path appears missing. These mismatches are unexpected and warrant caution.

ℹ Instruction Scope

Runtime instructions are mostly within the stated purpose (register an mcporter stdio server, read a local token JSON file on every call, and call Baidu API endpoints). But the SKILL.md uses absolute, environment-specific paths (e.g. /home/linuxbrew/.linuxbrew/bin/uv and /home/ubuntu/.openclaw/workspace/config/mcporter.json) and references a registration script that is not present in the bundle; it also instructs writing to a workspace mcporter.json which will change local tool config. Those assumptions may break or have side effects on different hosts.

✓ Install Mechanism

There is no remote install spec (no downloads from arbitrary URLs). The repository bundles Python server code and generated OpenAPI client code; nothing in the package attempts to fetch or execute remote installers. That lowers remote-install risk, but included code will be executed locally if the user follows SKILL.md steps—so inspect the code before running.

⚠ Credentials

Although the skill legitimately needs a Baidu access_token (the code reads BAIDU_NETDISK_TOKEN_FILE, BAIDU_NETDISK_ACCESS_TOKEN and BAIDU_NETDISK_DEFAULT_DIR), the registry metadata declares no required env vars or primary credential. This omission is misleading. Requiring an access token for cloud operations is proportionate to purpose, but the metadata mismatch and the fact the token lives in a local home path (~/.openclaw/credentials/baidudisk.json) are things a user should explicitly acknowledge before installation.

ℹ Persistence & Privilege

always:false and default autonomous invocation are reasonable. The SKILL.md instructs registering a stdio server entry into a mcporter config file under the workspace (modifying mcporter.json) — that is expected for integrating an MCP tool but is a local configuration change that affects how mcporter will invoke commands. The registration uses hard-coded command paths which may be inappropriate for some systems.

Version History

v1.0.0

Initial release of baidudisk-mcp - Provides Baidu Netdisk access through mcporter with an MCP stdio server, supporting hot-reload of token credentials. - Implements a wide set of official-aligned file management tools (aligned with Baidu API 2.0) and maintains compatibility with legacy tool aliases. - Supports batch operations with unified return envelope format for copy, move, and rename. - Credentials are read from a local JSON file on every call, enabling seamless token updates without restart. - Includes extensive examples for mcporter usage and recommended server setup steps.

Metadata

Slug baidudisk-mcp

Version 1.0.0

License —

All-time Installs 2

Active Installs 2

Total Versions 1

Frequently Asked Questions

What is Baidudisk Mcp?

Use Baidu Netdisk via mcporter + stdio MCP server with hot-reload token file credentials. Triggers when you need Baidu Netdisk operations (official 2.0 tools... It is an AI Agent Skill for Claude Code / OpenClaw, with 291 downloads so far.

How do I install Baidudisk Mcp?

Run "/install baidudisk-mcp" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Baidudisk Mcp free?

Yes, Baidudisk Mcp is completely free (open-source). You can download, install and use it at no cost.

Which platforms does Baidudisk Mcp support?

Baidudisk Mcp is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Baidudisk Mcp?

It is built and maintained by forevershu (@forevershu); the current version is v1.0.0.

More Skills

Baidudisk Mcp