Scholar Access 学术通道

通过中国教育网官方学术通道（scholar.edu.cn / scholar.work，即官方学术浏览器 scholartools 所使用的智能 DNS + 反向代理）访问 GitHub、HuggingFace、Docker Hub、Google Scholar、Wikipedia 等境外学术资源，让任何 Agent 无需翻墙即可合规访问。Routes blocked academic sites through the official CERNET "Scholar" reverse proxy used by the official scholartools academic-browser client — compliant, VPN-free access for any Agent.

Scholar Access — 学术资源浏览器协议模拟

The Chinese education network (scholar.edu.cn, backend scholar.work) runs a smart-DNS + reverse-proxy service so campus users can reach international academic resources. This is the official academic access channel used by the official academic-browser client scholartools (学术浏览器). scholartools normally installs a system-wide local DNS + root CA (needs root); this skill reproduces the same protocol as user-space tools that need no root, so any Agent can compliantly reach academic resources without a VPN (不翻墙).

Mechanism (reverse-engineered)

1. DNS-over-HTTPS at https://doh.scholar.work/dns-query maps a blocked foreign hostname onto a reverse-proxy farm in 205.164.50.0/24. Unsupported hostnames return REFUSED (0 answers).
2. The proxy at 205.164.50.x terminates TLS with a cert signed by Scholar Root CA v1 (bundled as scholar-root-ca.pem) and fetches the real site through the education network's international channel.
3. So: resolve via DoH → dial the 205.164.50.x IP with the original SNI/Host → trust the Scholar CA (or -k).

Bundled tools (all in this skill's directory)

File	Purpose
scholar-resolve.py <host>	Resolve a hostname via DoH; prints proxy IPs (205.164.50.x) or errors on REFUSED.
scholar-fetch.sh <url> [curl args...]	curl wrapper: auto-resolves and rewrites the connection via --resolve + bundled CA.
scholar-proxy.py [--listen 127.0.0.1:1080]	Local HTTPS CONNECT proxy for git / docker / pip / any HTTPS client (no root).
scholar-root-ca.pem	The Scholar Root CA v1 certificate.

SCHOLAR_CA env var overrides the CA path; SCHOLAR_PORT overrides the TLS port (default 443).

Quick start — fetch a URL

bash

cd "$(dirname "$0")"   # or use the skill's directory path
./scholar-fetch.sh https://github.com/torvalds/linux/archive/refs/heads/master.tar.gz -o linux.tar.gz
./scholar-fetch.sh https://huggingface.co/bert-base-uncased/resolve/main/config.json -o config.json
./scholar-fetch.sh https://raw.githubusercontent.com/octocat/Hello-World/master/README

Direct curl equivalent (if you prefer to inline it):

bash

IP=$(./scholar-resolve.py github.com | head -n1)   # e.g. 205.164.50.200
curl --resolve "github.com:443:$IP" --cacert scholar-root-ca.pem https://github.com/...
# or skip verification:  curl --resolve "github.com:443:$IP" -k https://github.com/...

git clone / git pull (no root)

Start the local proxy, then point git at it. The proxy keeps SNI correct, so plain https_proxy works:

bash

nohup ./scholar-proxy.py --listen 127.0.0.1:1080 > /tmp/scholar-proxy.log 2>&1 &
export https_proxy=http://127.0.0.1:1080 http_proxy=http://127.0.0.1:1080
GIT_SSL_NO_VERIFY=1 git clone https://github.com/<owner>/<repo>.git
# When finished:
unset https_proxy http_proxy
pkill -f scholar-proxy.py   # or keep it running for the session

git clone, git fetch, git pull and git ls-remote all work through this proxy for github.com.

docker pull

registry-1.docker.io and hub.docker.com are proxied. Route the docker daemon through the local proxy (daemon-level, needs a docker daemon restart on Linux) or use docker pull with a proxy-configured daemon:

bash

# export these before starting dockerd, or set in /etc/systemd/system/docker.service.d/http-proxy.conf
export HTTPS_PROXY=http://127.0.0.1:1080 HTTP_PROXY=http://127.0.0.1:1080
docker pull nginx

Verified proxied domains

205.164.50.x is returned for (non-exhaustive): github.com, raw.githubusercontent.com, huggingface.co, hf.co, cdn-lfs.hf.co (model files), hub.docker.com, registry-1.docker.io, scholar.google.com, www.google.com, www.google.com.hk, www.wikipedia.org, en.wikipedia.org, www.anthropic.com, www.perplexity.ai, chatgpt.com, biorxiv.org.

Not proxied (DoH returns REFUSED — do not expect these to work): api.github.com, codeload.github.com, objects.githubusercontent.com, gist.githubusercontent.com, api.anthropic.com, pypi.org, stackoverflow.com.

Implication: GitHub web UI and raw file access work, but the GitHub REST API and release-asset downloads (objects.githubusercontent.com) do not go through this proxy. For GitHub release assets, fetch the repo page and use a direct raw.githubusercontent.com or github.com asset URL when possible.

When the proxy is not needed

If the target hostname is reachable directly (or the DoH returns a non-205.164.50.x IP, e.g. gitlab.com), skip the proxy and connect normally.

Diagnostics

bash

./scholar-resolve.py github.com          # expect 205.164.50.x
./scholar-resolve.py api.github.com      # expect ERROR (REFUSED)
curl -sk --resolve github.com:443:205.164.50.200 https://github.com/ | head -c 200