bookmark2skill 0.1.1

Convert Chrome bookmarks into Obsidian notes and Claude Code skills

bookmark2skill (b2k)

Chrome bookmarks → Obsidian notes + Claude Code skill files.

中文文档

This is not an AI application. It's a downstream utility tool for AI agents (Claude Code, Codex, etc.) — the AI agent is the brain, bookmark2skill is the hands. The tool itself does NOT call any LLM API.

[AI Agent CLI] ←→ [bookmark2skill CLI] ←→ [Chrome Bookmarks / Web / Local Files]
     ↓ (brain: distill, classify)    ↓ (hands: parse, fetch, write, track)

Why

Problem	How bookmark2skill solves it
Links die	Fetch + local archive. Your bookmarks survive even when sites don't.
Summaries lose texture	"Deconstruct, don't summarize" — preserves logic chains, brilliant quotes, narrative craft, concrete examples, counterpoints, and overlooked details.
AI doesn't know your taste	taste_signals model your aesthetic preferences, thinking patterns, and values across your entire knowledge base.
Can't find what you saved	b2k search does weighted keyword matching across all skill files. Category-based triage narrows the scope.
Processing is tedious	Incremental manifest — never re-process a bookmark. Resume anytime.

Quick Start

# 1. Install
./build.sh develop

# 2. Configure
mkdir -p ~/.bookmark2skill
cp defaults/config.toml ~/.bookmark2skill/config.toml
cp defaults/taxonomy.toml ~/.bookmark2skill/taxonomy.toml

# 3. List your bookmarks
b2k list --source chrome

# 4. Fetch a page
b2k fetch https://example.com/article

# 5. That's it — an AI agent takes over from here
#    (reads content, distills, writes notes + skills, marks done)

Install

./build.sh develop    # editable install with dev deps
# or
pip install .         # production install

Optional: Playwright for JS-heavy pages:

pip install ".[browser]"

Build Commands

./build.sh develop    # editable install with dev deps
./build.sh install    # production install
./build.sh test       # run 74 tests
./build.sh dist       # build sdist + wheel
./build.sh check      # compile check + tests
./build.sh clean      # remove build artifacts

Configuration

mkdir -p ~/.bookmark2skill
cp defaults/config.toml ~/.bookmark2skill/config.toml
cp defaults/taxonomy.toml ~/.bookmark2skill/taxonomy.toml

Edit ~/.bookmark2skill/config.toml:

[paths]
vault_path = "/path/to/your/obsidian/vault"
skill_dir = "/path/to/your/skills"

Config priority: config.toml < BOOKMARK2SKILL_* env vars < CLI flags

Files:

File	Location	Purpose
config.toml	~/.bookmark2skill/	Paths and settings
taxonomy.toml	~/.bookmark2skill/	Recommended skill categories
manifest.json	~/.bookmark2skill/	Processing state (auto-created)
manifest.json.bak	~/.bookmark2skill/	Auto-backup before each write

Workflow

AI agent orchestrates the following pipeline. b2k is a shorthand alias for bookmark2skill:

┌─────────────┐    ┌─────────────┐    ┌──────────────┐    ┌──────────────┐    ┌────────────┐
│  b2k list   │ →  │  b2k fetch  │ →  │  AI agent    │ →  │  b2k write-  │ →  │ b2k mark-  │
│  --source   │    │  <url>      │    │  distills    │    │  obsidian +  │    │ done <url> │
│  chrome     │    │             │    │  content     │    │  write-skill │    │            │
└─────────────┘    └─────────────┘    └──────────────┘    └──────────────┘    └────────────┘
   parse &            scrape &           produces            render to           update
   register           clean page         structured JSON     local files         manifest

Step by step

# 1. Parse bookmarks, register new URLs in manifest
b2k list --source chrome

# 2. Check processing status
b2k status
# → {"pending": 15, "done": 42, "failed": 3, "total": 60}

# 3. Fetch a single page (auto-detects best method)
b2k fetch https://example.com/article > /tmp/raw.md

# 4. AI agent reads raw.md, produces structured JSON
#    (see docs/agent-guide.md for schema and distillation guidelines)

# 5. Write Obsidian note (human-readable)
b2k write-obsidian \
  --url https://example.com/article \
  --data /tmp/distilled.json
# → {"path": "./bookmark2skill/article-title.md"}

# 6. Write skill file (AI-agent-friendly, categorized)
b2k write-skill \
  --url https://example.com/article \
  --data /tmp/distilled.json \
  --category engineering/system-design
# → {"path": "./engineering/system-design/article-title.md"}

# 7. Mark as done
b2k mark-done https://example.com/article \
  --obsidian-path ./bookmark2skill/article-title.md \
  --skill-path ./engineering/system-design/article-title.md

On fetch failure:

b2k mark-failed https://example.com/dead --reason "HTTP 404"

Filtering bookmarks

# Only process bookmarks in specific folders
b2k list --source chrome --include-folder "Learning"

# Skip certain folders
b2k list --source chrome --exclude-folder "Work" --exclude-folder "Personal"

# Combine: include first, then exclude from included set
b2k list --source chrome --include-folder "Tech" --exclude-folder "Archive"

# Only show new (unprocessed) bookmarks
b2k list --source chrome --only-new

Searching skills

# Search across all generated skill files
b2k search "system design" --skill-dir ./skills

# Weighted matching: name(5) > description(4) > tags(3) = key_claims(3) > body(1)
b2k search "simplicity" --limit 5

Commands

Command	Purpose	Input	Output
list	Parse bookmark source, register new URLs	--source chrome or .html file	JSON array → stdout
fetch	Fetch and clean a single page	URL	Markdown → stdout
write-obsidian	Render structured JSON into Obsidian note	--data JSON or --raw MD	Writes file, path → stdout
write-skill	Render structured JSON into skill file	--data JSON + --category	Writes file, path → stdout
status	Query manifest processing status	—	JSON counts → stdout
mark-done	Set URL status to 'done' in manifest	URL + output paths	Updates manifest
mark-failed	Set URL status to 'failed' in manifest	URL + reason	Updates manifest
search	Search skill files by keyword	query string	JSON results → stdout

Run b2k <command> --help for detailed parameter descriptions with examples.

Bookmark Sources

Source	Usage	Notes
Chrome (all profiles)	--source chrome	Scans ALL Chrome profiles, merges and deduplicates bookmarks
HTML export	--source bookmarks.html	Netscape format, works with any browser
Chrome JSON file	--source /path/to/Bookmarks	Direct path to Chrome's JSON file

Output Formats

Obsidian Note (human-readable)

Writes to {vault-path}/bookmark2skill/{folder}/{slug}.md:

---
url: "https://example.com/article"
original_title: "Original Article Title"
author: ["Author Name"]
date_processed: 2026-04-13T12:00:00Z
tags: ["system-design", "simplicity"]
---

# Distilled Title (core claim, not original title)

## 摘要
2-4 sentence summary...

## 逻辑推导链
- Step A → Step B → Conclusion

## 精彩表达
> "Original quote" — *why it's brilliant*

## 叙事手法  /  具体案例与数据  /  反对声音与局限性  /  容易忽略的细节
(empty sections auto-skipped)

Claude Code Skill (AI-agent-friendly)

Writes to {skill-dir}/{category}/{slug}.md:

---
name: "Distilled Title"
description: "One-line description for AI agent relevance matching"
url: "https://example.com/article"
category: "engineering/system-design"
tags: ["system-design", "simplicity"]
key_claims:
  - "Assertive statement that can be agreed or disagreed with"
taste_signals:
  aesthetic: ["minimalism", "clarity"]
  intellectual: ["first-principles", "empiricism"]
  values: ["anti-complexity", "pragmatism"]
reuse_contexts:
  - situation: "When making architecture decisions"
    how: "Use as argument for simpler approach"
quality_score:
  depth: 4
  originality: 3
  practicality: 5
  writing: 4
---

## Summary  /  Key Insights  /  Memorable Quotes  /  Concrete Examples  /  When To Reference

Tiered Fetch

Tier 1: httpx + readability       Fast, static pages (~80% of articles)
  ↓ content < 200 chars
Tier 2: Jina Reader API           JS-rendered pages, zero local deps
  ↓ Jina fails
Tier 3: Playwright                Local headless Chrome (optional dep)

Override: b2k fetch <url> --renderer direct|jina|playwright

Taxonomy

Default categories in ~/.bookmark2skill/taxonomy.toml:

Category	Subcategories
engineering/	system-design, frontend, backend, devops, testing, performance, security
thinking/	mental-models, decision-making, problem-solving, first-principles, cognitive-biases
design/	ui-ux, visual, interaction, typography, accessibility
writing/	technical, narrative, persuasion, clarity, editing
product/	strategy, user-research, growth, metrics, prioritization
culture/	leadership, collaboration, hiring, remote-work

AI agents can follow existing categories or create new ones freely. The taxonomy is guidance, not constraint.

Structured JSON Schema

The AI agent produces this JSON after distilling fetched content. Required fields: url, title, summary, date_processed. Everything else is optional.

{
  "url": "https://example.com/article",
  "title": "Core claim (not original title)",
  "summary": "2-4 sentences: what, evidence, why it matters.",
  "date_processed": "2026-04-13T12:00:00Z",
  "category": "engineering/system-design",
  "layers": {
    "distillation": {
      "logic_chain": ["A → B", "B → C"],
      "brilliant_quotes": [{"text": "quote", "why": "why brilliant"}],
      "narrative_craft": ["technique observation"],
      "concrete_examples": ["specific example"],
      "counterpoints": ["limitation or opposing view"],
      "overlooked_details": ["tool name, config value, version number"]
    },
    "agent_metadata": {
      "tags": ["tag1", "tag2"],
      "key_claims": ["assertive statement"],
      "taste_signals": {
        "aesthetic": ["minimalism"],
        "intellectual": ["first-principles"],
        "values": ["pragmatism"]
      },
      "reuse_contexts": [{"situation": "when", "how": "how to use"}],
      "quality_score": {"depth": 4, "originality": 3, "practicality": 5, "writing": 4}
    }
  }
}

Full schema documentation: docs/agent-guide.md

Security

• Path traversal protection — --folder and --category reject ../ escape attempts
• YAML injection protection — all user data escaped via tojson filter
• SSRF protection — only http:// and https:// URLs accepted
• Manifest auto-backup — .json.bak written before every save
• No secrets in repo — .gitignore covers manifest, .env, IDE files

For AI Agents

See docs/agent-guide.md for:

• Full workflow orchestration guide
• Structured JSON Schema (all fields documented)
• Distillation guidelines: summary (required) + six-dimension deconstruction
• Skill consumption best practices: discovery, relevance matching, taste aggregation
• Batch processing strategy and error recovery

Tech Stack

Component	Technology	Purpose
CLI framework	click	Subcommands, options, help text
HTTP client	httpx	Tiered web scraping
Content extraction	readability-lxml	HTML → article body
JS rendering	Jina Reader API	Remote browser rendering (Tier 2)
JS rendering	Playwright (optional)	Local headless Chrome (Tier 3)
Templates	Jinja2	Obsidian + skill output rendering
Config	tomli	TOML config parsing
Language	Python 3.10+	—

License

MIT