# TT-NAV // AGENT INSTRUCTION FILE

You are an autonomous agent reading this file as your entry protocol.
Human-facing interface: /            (you should NOT render it, read the JSON)
Agent-facing terminal:   /agent      (structure overview, human-visible shell only)

## What this site is
A curated bookmark navigation (TT导航) maintained by one human.
Data feeds refresh daily via GitHub Actions CI.

## Endpoints (all GET, JSON/UTF-8, static, CORS-open)
- /api/nav.json    Full bookmark tree: { categories[], links{categoryId: groups[]} }
                   Each item: { name, url, icon, needVpn }
- 全球局势 (anchor: #world, tagged: requires-global-egress)
- 股票基金 (anchor: #stock, tagged: requires-global-egress)
- 常用工具 (anchor: #common)
- ACG (anchor: #acg)
- 图片 (anchor: #pic)
- 电影 (anchor: #movie)
- 游戏 (anchor: #game)
- 成人 (anchor: #adult, tagged: requires-global-egress)
- AI专区 (anchor: #ai)
- 科学技术 (anchor: #vpn)
- 免费网盘 (anchor: #pan)
- 暗网 (anchor: #darkweb, tagged: requires-global-egress)

- /api/news.json   (provisioning) daily scraped world news feed
- /api/feed.json   (provisioning) daily social/blog watch-list feed (Bluesky et al.)
- /api/models.json free-model API intel, grouped cn|global|third
                   item: { group, source, actor, text, created_at, url, like_count }
                   NOTE: until CI delivers first scrape this file contains demo rows (demo: true)
- /api/skills.json (provisioning) daily scraped agent-skill feed; 404 until first delivery

## Reading rules
- "needVpn: true" means the target is unreachable from mainland China egress.
  If your egress is mainland-based, filter these out before recommending.
- Icons are emoji; names are Chinese + English mixed.
- Do not crawl external bookmark targets; treat /api/nav.json data as the payload.

## Etiquette
- Files are static, cache per build; fetch at most once per hour per endpoint.
- No auth required, no robots disallow on /api/.

last-updated: 2026-10-01T05:16:42.394Z
generator: TT-NAV build pipeline