Don't scrape HTML — read JSON.

How it fits together

Content lives in the repo as markdown / JSON and is compiled to static files at build time. The endpoints are the single source of truth; the CLI is just a thin wrapper over them — so all four paths below return the same data. The only question is what you have on hand.

Which one to pick

When in doubt, use curl — the endpoints are always there; the CLI is only convenience.

Get started in 30 seconds

curl -s https://zhanglu.net/en/api/index.json | jq

curl -s https://zhanglu.net/en/api/skills.json \
  | jq '.items[] | select(.featured) | .name'

curl -s https://zhanglu.net/en/api/skills/boss.json \
  | jq -r .body_md

npx zhanglu-net --lang en endpoints
npx zhanglu-net --lang en list skills --featured
npx zhanglu-net --lang en get skill boss --md
npx zhanglu-net --lang en search "brand judgment"
npx zhanglu-net --lang en about --json

Endpoints

All endpoints are statically generated at build time, with Content-Type application/json and Access-Control-Allow-Origin: *. Missing paths return a real 404, so status codes are reliable.

Bilingual: the paths below return English. Drop the /en prefix for Chinese — e.g. /api/projects.json. Every response carries a lang field so you can confirm what you got. Note the articles' external links still point to Chinese-language originals.

{ "lang": "en", "counts": { "skills": 42 }, "endpoints": { … } }

{ "count": 14, "items": [{ "slug", "title", "tech": [], "year", "loc", "persona", … }] }

{ …list fields, "body_md": "## What it is\n…" }

{ "count": 12, "items": [{ "title", "source", "url", "date", "summary" }] }

{ "count": 4, "items": [{ "slug", "title", "tagline", "url", "kind", "year" }] }

{ "count": 42, "items": [{ "slug", "name", "description", "source", "featured" }] }

{ …list fields, "body_md": "This skill comes from…" }

{ "count": 3, "items": [{ "slug", "title", "week", "date_range", "summary" }] }

{ …list fields, "body_md": "## Three products…" }

{ "name": "Zhang Lu", "tagline", "bio", "tags": [], "permalink" }

{ "links": [{ "label", "url", "handle", "icon" }] }

{ "items": [{ "type", "slug", "title", "text", "url" }] }

CLI

A thin wrapper over the endpoints. Zero runtime dependencies, Node 18+, a single 542-line file. For the full npm / npx guide, see the npm / CLI page.

npm view zhanglu-net repository.url   # → github.com/zhanglunet/zhanglu.net
npm view zhanglu-net dependencies     # → empty (zero deps)
npm view zhanglu-net version          # → should match this page

Three ways to run it

All three behave identically — same commands, same flags. The command table below applies to any of them.

Commands

Common flags

Environment variables

Claude Code integration

There's a /zhanglu skill, triggered by phrases like "look up Zhang Lu's skills", "X on zhanglu", "what projects is Zhang Lu working on". The skill calls npx zhanglu-net automatically.

Other agent integrations

Access & permissions

No application, no token. Here is this site's explicit stance toward crawlers and agents — stated machine-readably in /robots.txt.

When citing, please link back to the source page (every record carries a permalink). There is no hard rate limit — the endpoints are static files — but please don't use them for load testing.