open-bookmark

Spec-driven browser bookmark management. Define your rules first, then execute.

Features

• Scan: Read Chrome bookmarks and analyze patterns
• Analyze: Identify domains, folders, keywords, and URL patterns
• Profile: Infer user role, tech stack, interests, and language
• Rules: Generate classification rules based on analysis
• Preview: Show what rules will do before applying
• Apply: Execute rules to reorganize bookmarks
• Verify: Check results after applying rules
• Rollback: Restore from backup if needed
• AI-Enhanced: Multi-agent pipeline for intelligent classification

Install

npm install -g open-bookmark

Quick Start

# Initialize: scan bookmarks, analyze, generate rules
open-bookmark init

# Preview rule execution effects
open-bookmark preview

# Apply rules to bookmarks
open-bookmark apply

# View and modify user profile
open-bookmark config --show

# Generate incremental rules for new bookmarks
open-bookmark propose

# Install SKILL.md for AI platforms
open-bookmark skill install

AI-Enhanced Features

Multi-Agent Pipeline

Run the enhanced multi-agent pipeline for intelligent bookmark classification:

# Run enhanced pipeline
open-bookmark pipeline --enhanced

# Save results to JSON
open-bookmark pipeline --enhanced --json --output results.json

The pipeline uses three agents:

1. Scanner Agent: Extracts features from bookmarks (domains, URLs, paths)
2. Analyzer Agent: Performs semantic analysis with hierarchical classification
3. Classifier Agent: Generates classification rules with dynamic priority

Hierarchical Classification

The system supports hierarchical classification with:

• Domain-based rules: Highest priority (0.95 confidence)
• Keyword-based rules: Medium priority (0.5-0.9 confidence)
• URL path patterns: Lower priority (0.7-0.8 confidence)
• Folder structure: Lowest priority (0.6 confidence)

Example hierarchy:

DevOps
├── DevOps/Containers (Docker, Podman)
├── DevOps/Orchestration (Kubernetes, Helm)
├── DevOps/CI-CD (Jenkins, GitLab CI)
├── DevOps/Monitoring (Prometheus, Grafana)
└── DevOps/IaC (Terraform, Ansible)

Quality Evaluation

Evaluate classification quality:

# Evaluate with enhanced evaluator
open-bookmark evaluate --enhanced

# Output as JSON
open-bookmark evaluate --enhanced --json

The evaluator reports:

• Coverage percentage
• Category distribution
• Hierarchy depth
• Quality score (0-100)

User Feedback

Collect and manage user feedback:

# Add feedback
open-bookmark feedback --add "id:name:from:to:reason"

# List pending feedback
open-bookmark feedback --list

# Generate feedback report
open-bookmark feedback --report

Rule Optimization

Optimize rules based on feedback:

# Run optimization analysis
open-bookmark optimize

# Apply optimizations
open-bookmark optimize --apply

AI-Enhanced Workflow

For best results, use the complete AI-enhanced workflow:

# Step 1: Initialize and scan bookmarks
open-bookmark init

# Step 2: Run enhanced pipeline
open-bookmark pipeline --enhanced

# Step 3: Evaluate classification quality
open-bookmark evaluate --enhanced

# Step 4: Provide feedback on misclassified bookmarks
open-bookmark feedback --add "id:name:from:to:reason"

# Step 5: Optimize rules based on feedback
open-bookmark optimize --apply

# Step 6: Preview and apply
open-bookmark preview
open-bookmark apply

Commands

| Command | Description | |---------|-------------| | init | Initialize open-bookmark: scan bookmarks, analyze, generate rules | | status | Show current status and next steps | | preview | Preview rule execution effects | | apply | Apply rules to bookmarks | | verify | Verify results after apply | | rollback | Rollback to last backup | | analyze | Deep analysis of bookmarks (read-only) | | config | View and modify user profile | | propose | Scan new bookmarks and generate incremental rules | | skill | Generate and install SKILL.md for AI platforms | | prepare | Prepare bookmark data for AI analysis | | rules | Convert AI-generated tags to rules | | stats | Output bookmark statistics for AI | | pipeline | Run multi-agent analysis pipeline | | evaluate | Evaluate classification quality | | feedback | Manage user feedback | | optimize | Run optimization cycle |

Knowledge Graph

Build a knowledge graph of your bookmarks and export it as standard JSON for analysis or import into graph tools (e.g. Gephi, Cytoscape):

# Build the graph
open-bookmark graph init

# Query, stats, and learning paths
open-bookmark graph query <keyword>
open-bookmark graph stats
open-bookmark graph tour

# Export the graph as JSON
open-bookmark graph export > bookmark-graph.json

Rule Format

Rules are stored in YAML format with match conditions:

- name: kubernetes-生态
  match:
    domain:
      - kubernetes.io
      - "*.kubernetes.io"
      - helm.sh
    title_contains:
      - k8s
      - kubectl
  action: move
  target: DevOps/Kubernetes
  reason: "Kubernetes and cloud-native tools"
  source: generated

Match Conditions

• domain — Domain matching (supports wildcards)
• url_regex — URL regex matching
• title_contains — Title keyword matching
• title_exclude — Title exclusion
• folder_path — Exact folder path
• folder_prefix — Folder prefix matching
• match_all — Catch-all rule

Actions

• move — Move bookmark to target folder
• skip — Skip bookmark (protected)
• analyze — Mark for analysis

AI Platform Integration

open-bookmark supports SKILL.md generation for major AI code tools:

• Claude Code
• Cursor
• OpenCode
• Codex
• Gemini CLI
• GitHub Copilot
• Cline
• Kimi

By default the skill is installed to your user-global directory so it is available from any project. Pass --project to install into the current project instead.

Install SKILL.md:

open-bookmark skill install            # user-global (default)
open-bookmark skill install --project  # current project

Generate SKILL.md locally:

open-bookmark skill generate

Documentation

• User Guide — Complete user guide
• API Reference — API documentation
• Developer Guide — Contributing guide

License

MIT