DejaViewed

"You've saved this before."

Turn your saved posts and bookmarks from any platform into a searchable, curated dark-mode catalog site with an auto-generated Action Plan — powered by Claude Code.

You scroll, you tap save, you swear you'll come back. You don't. DejaViewed hands the whole pile to Claude: every save gets read, classified, tiered, and — for the ones worth the dig — turned into a real deep-dive guide with the links the creator wouldn't give you. Then it tells you what to actually do with all of it.

Live demo: dejaviewed.dev

Install

# Full plugin install
git clone git@github.com:Gabicus/dejaviewed.git ~/.claude/plugins/dejaviewed

# Or skill-only install
git clone git@github.com:Gabicus/dejaviewed.git /tmp/dv && cp -r /tmp/dv/skills/dejaviewed ~/.claude/skills/

Use

/dejaviewed

Or with arguments:

/dejaviewed --urls saved_urls.json --title "My AI Bookmarks" --handle "@myhandle"

Supported Sources

Source	Auth Required	Status
Instagram Saves	Chrome profile copy	✅ Built
Chrome Bookmarks	None (local file)	✅ Built
Firefox Bookmarks	None (local file)	✅ Built
Edge Bookmarks	None (local file)	✅ Built
TikTok Saves	Chrome profile copy	🔜 Planned
Twitter/X Bookmarks	API bearer token	🔜 Planned
Pinterest Boards	Chrome profile / API	🔜 Planned
Reddit Saved	OAuth token	🔜 Planned
YouTube Watch Later	Chrome profile / API	🔜 Planned

What It Does

Phase	What happens
Ingest	Pull saves/bookmarks from any supported source into a unified format
Classify	Claude reads each item, writes a "Subject — why I saved it" title, summary, type, tools/repos mentioned
Links	Extracts every outbound URL, GitHub repo, tool website, @mention
Tier	S/A/B/C ranking based on how actionable and substantive each item is
Deep-Dive	Markdown guides for the top-tier items — the research the creator didn't give you
Thumbnails	Downloads images locally (platform CDN URLs expire — never hotlink)
DejaViewed Page	Auto-generates "what to DO with all this" — clone commands, install commands, guides to read, techniques to try
Render	Static HTML: masonry layout, category filters, creator bar chart, clickable links, search, DejaViewed summary page

What You Get

A static HTML site with:

• Masonry 3-column card layout — cards size to content, not forced grids
• Dark mono theme — SF Mono / Fira Code, dark panels, gradient accents
• Sticky sidebar — multi-select category filters, tier pills, full-text search
• Creator bar chart — proportional bars colored by tier, click a segment to jump to that card
• Clickable links on every card — repos, tools, websites, @handles extracted from captions
• Thumbnails — downloaded locally, not hotlinked
• Deep-dive guide pages — full markdown with confidence badges
• Per-collection pages — if you have multiple saved folders
• DejaViewed summary page — the killer feature, first tab:
  • Hero layout matching catalog pages (title + WHY + BANs left, install/sources sidebar right)
  • Section jump pills linking to each action category
  • Clone These Repos — S/A tier repos with git clone commands (click to copy)
  • Install These Tools — pip/npm/brew one-liners
  • Read These Guides — deep-dives we wrote, with confidence badges
  • Try These Techniques — workflows you saved but haven't tried
  • Bookmark These Platforms — services worth keeping
  • Design & Art Resources — archives, collections, inspiration
  • Save Profile — witty personality assessment of your hoarding habits
• Grouped navigation — DEJAVIEWED → ALL → source labels with sub-page pills (e.g., Instagram → AI1/AI2/AI3/AI4)

Browser Bookmarks (No Auth Needed)

The easiest way to start — no cookies or API tokens required:

# Extract Chrome bookmarks
python3 adapters/chrome_bookmarks.py --out data/chrome_bookmarks.jsonl

# Extract Firefox bookmarks
python3 adapters/firefox_bookmarks.py --out data/firefox_bookmarks.jsonl

# Merge all sources (dedup by URL)
python3 adapters/merge_sources.py \
  --sources data/catalog.jsonl data/chrome_bookmarks.jsonl \
  --out data/catalog_merged.jsonl --dedup-by url

Cross-source dedup: if the same URL appears from Chrome AND Instagram, the merger keeps the richer record and adds sources: ["instagram", "chrome"].

Prerequisites

For Instagram scraping:

1. Instagram saved post URLs as JSON: {"collection":"name","count":N,"urls":[...]}
2. Copied Chrome profile with active IG session at .profile-copy/Default/Cookies
3. Python 3.10+ with requests and browser_cookie3

For browser bookmarks only:

1. Python 3.10+ (no additional packages needed)
2. Chrome, Firefox, or Edge installed with bookmarks

Example Prompt

See skills/dejaviewed/references/example-prompt.md for a ready-to-copy template.

Security

• Cookie values are NEVER printed, logged, or written to any output
• All card rendering via DOM createElement — no innerHTML (DOMPurify for guide markdown only)
• Platform CDN thumbnails downloaded locally, not hotlinked
• Browser bookmark adapters ONLY read bookmarks — never history, passwords, or cookies

Credits

Built by @6ab3 with Claude Code.

License

MIT