Skill Seekers: конвертер документации в AI-скиллы

10.03.2026обновлено 6 апр

Есть library с отличной документацией, но нет готового скилла? Skill Seekers от Yusuf Karaaslan автоматически конвертирует документацию, GitHub репозитории и PDF в ready-to-use Claude AI skills.

Documentation → Skills Pipeline

Skill Seekers решает fundamental проблему: 99% полезных tools и libraries не имеют ready-made skills для AI-агентов. Manual создание skills — slow и error-prone process.

Automatic pipeline:

  1. SCRAPE — documentation, GitHub repos, PDFs
  2. ANALYZE — extract knowledge patterns
  3. ENHANCE — AI-powered skill optimization
  4. PACKAGE — ready-to-use skill files

Supported sources:

  • Documentation websites (any structure)
  • GitHub repositories (code + README + wiki)
  • PDF documents (technical guides, manuals)
  • Word documents (.docx)
  • YouTube videos (transcription → skill)
  • Multi-source unified scraping

Установка и конфигурация

pip install skill-seekers

Initial configuration:

# Настройка API ключей и GitHub tokens
skill-seekers config

# Interactive setup:
# - GitHub Personal Access Token
# - AI enhancement API keys (optional)
# - Output preferences
# - Quality thresholds

Архитектура:

Skill Seekers Pipeline:
Source → Scraper → Analyzer → Enhancer → Packager → Deploy

Основные команды

Auto-detection (рекомендуется):

# Smart detection source type
skill-seekers create --source "https://docs.react.dev" --name "react-guide"
skill-seekers create --source "microsoft/TypeScript" --name "typescript"  
skill-seekers create --source "python-guide.pdf" --name "python-guide"

Documentation scraping:

# Scrape documentation website
skill-seekers scrape --config configs/react.json

# Custom configuration
skill-seekers scrape \
  --url "https://docs.nextjs.org" \
  --name "nextjs-guide" \
  --max-pages 100 \
  --depth 3

GitHub repository analysis:

# Scrape GitHub repository
skill-seekers github --repo microsoft/TypeScript --name typescript

# Include specific patterns  
skill-seekers github \
  --repo facebook/react \
  --name react-internals \
  --include "*.md,*.ts,*.js" \
  --exclude "test/*,example/*"

Configuration Files

Example: React documentation config

{
  "name": "react-docs",
  "base_url": "https://react.dev",
  "start_urls": [
    "https://react.dev/learn",
    "https://react.dev/reference"
  ],
  "allowed_domains": ["react.dev"],
  "exclude_patterns": [
    "/blog/",
    "/versions/"
  ],
  "max_pages": 200,
  "depth_limit": 4,
  "extract_code": true,
  "include_examples": true,
  "quality_threshold": 0.7
}

GitHub repository config:

{
  "repo": "microsoft/TypeScript",
  "name": "typescript-compiler",
  "branches": ["main"],
  "include_paths": [
    "src/compiler/",
    "src/services/",
    "lib/"
  ],
  "file_patterns": ["*.ts", "*.md"],
  "max_file_size": "1MB",
  "extract_comments": true,
  "include_tests": false
}

Multi-Source Unified Scraping

Complete knowledge extraction:

# Unified: docs + GitHub + PDF
skill-seekers unified --config configs/complete_react.json

Unified config example:

{
  "name": "react-complete",
  "sources": [
    {
      "type": "docs",
      "url": "https://react.dev",
      "max_pages": 100
    },
    {
      "type": "github", 
      "repo": "facebook/react",
      "include": "packages/react/src/"
    },
    {
      "type": "pdf",
      "file": "react-patterns-guide.pdf"
    }
  ],
  "merge_strategy": "weighted",
  "deduplicate": true
}

AI-Powered Enhancement

Auto-enhancement:

# AI-powered optimization
skill-seekers enhance output/react/

# Background processing (для больших skills)
skill-seekers enhance output/react/ --daemon

# Check status
skill-seekers enhance-status

Enhancement features:

  • Structure optimization** — reorganize content for better flow
  • Example generation** — create practical usage examples
  • Cross-referencing** — link related concepts
  • Quality scoring** — assess completeness and accuracy
  • Template generation** — standardized skill format

Enhancement workflow presets:

# Show available presets
skill-seekers workflows list

# Use specific workflow
skill-seekers enhance output/react/ --workflow "documentation-to-skill"

# Custom workflow
skill-seekers enhance output/react/ --workflow custom.json

Specialized Extraction

PDF extraction:

# Technical documentation PDF
skill-seekers pdf --file "kubernetes-guide.pdf" --name "k8s-ops"

# With OCR для scanned PDFs
skill-seekers pdf \
  --file "legacy-manual.pdf" \
  --name "legacy-system" \
  --ocr \
  --lang en

Video content extraction:

# YouTube technical talks
skill-seekers video \
  --url "https://youtube.com/watch?v=tech-talk-id" \
  --name "advanced-react" \
  --transcript-only

# Local video files
skill-seekers video \
  --file "conference-talk.mp4" \
  --name "microservices-patterns"

Word document extraction:

# Corporate documentation
skill-seekers word \
  --file "api-guidelines.docx" \
  --name "api-standards" \
  --extract-tables \
  --preserve-formatting

Quality Assessment

Built-in quality scoring:

# Assess skill quality
skill-seekers quality output/react/

# Detailed analysis report
skill-seekers quality output/react/ --detailed --report quality-report.html

Quality metrics:

  • Completeness** — coverage of source material
  • Coherence** — logical structure and flow
  • Accuracy** — factual correctness
  • Usability** — practical applicability
  • Examples** — quantity and quality of usage examples

Quality thresholds:

{
  "quality_thresholds": {
    "completeness": 0.8,
    "coherence": 0.7, 
    "accuracy": 0.9,
    "usability": 0.8,
    "examples": 0.6
  }
}

Advanced Features

Incremental updates:

# Update existing skill без full rescrape
skill-seekers update output/react/ --check-changes

# Smart update strategy
skill-seekers update output/react/ \
  --strategy "delta" \
  --preserve-customizations

Multi-language support:

# Documentation в multiple languages
skill-seekers multilang \
  --base-url "https://docs.example.com" \
  --languages "en,es,fr,de" \
  --name "multilang-guide"

Stream processing (для больших sources):

# Stream large repositories chunk-by-chunk
skill-seekers stream \
  --repo "tensorflow/tensorflow" \
  --chunk-size 1000 \
  --parallel 4

Resume interrupted jobs:

# Resume после network interruption
skill-seekers resume --job-id "react-docs-20240310"

# Show active jobs
skill-seekers jobs list

Complete Workflow Example

Example: Create Kubernetes skill

Step 1: Estimate scope

skill-seekers estimate --url "https://kubernetes.io/docs" --depth 3
# Output: ~450 pages, estimated 2-3 hours

Step 2: Multi-source scraping

skill-seekers unified --config k8s-config.json

k8s-config.json:

{
  "name": "kubernetes-ops",
  "sources": [
    {
      "type": "docs",
      "url": "https://kubernetes.io/docs",
      "max_pages": 300,
      "focus_sections": ["concepts", "tutorials", "reference"]
    },
    {
      "type": "github",
      "repo": "kubernetes/kubernetes", 
      "include": "docs/", 
      "exclude": "vendor/"
    },
    {
      "type": "pdf",
      "file": "k8s-best-practices.pdf"
    }
  ]
}

Step 3: AI enhancement

skill-seekers enhance output/kubernetes-ops/ --workflow "ops-guide"

Step 4: Quality check

skill-seekers quality output/kubernetes-ops/ --threshold 0.8

Step 5: Package and deploy

skill-seekers package output/kubernetes-ops/
skill-seekers install-agent output/kubernetes-ops.skill

Integration с Claude Code

Auto-install to agent directories:

# Install to Claude Code skills directory
skill-seekers install-agent output/react.skill

# Bulk install multiple skills
skill-seekers install-agent output/*.skill --directory /path/to/claude/skills

Testing generated skills:

# Extract test examples from source
skill-seekers extract-test-examples output/react/

# Generate validation prompts
skill-seekers test-gen output/react/ --count 10

Best Practices

For Documentation Scraping:

  • Start с small scope, expand gradually
  • Use --estimate before full scrape
  • Configure appropriate depth limits
  • Exclude irrelevant sections (blog, changelog)
  • Monitor rate limits

For GitHub Repositories:

  • Focus на documentation и core source
  • Exclude test files unless specifically needed
  • Use file pattern filtering
  • Respect repository size limits
  • Include README и wiki content

For Quality Enhancement:

  • Always run AI enhancement на final output
  • Use workflow presets для consistent results
  • Set appropriate quality thresholds
  • Review generated examples for accuracy
  • Validate cross-references

For Maintenance:

  • Set up incremental update schedules
  • Monitor source changes
  • Preserve manual customizations
  • Version control skill changes
  • Regular quality assessments

Advanced Use Cases

Enterprise documentation pipeline:

# Corporate knowledge extraction
skill-seekers unified --config enterprise-config.json
# Sources: internal docs + GitHub enterprise + Confluence + PDFs

# Automated skill generation for internal tools
skill-seekers github --org company-internal --bulk --auto-enhance

Open source project onboarding:

# Generate comprehensive project guide
skill-seekers create --source "https://github.com/apache/kafka" --complete
# Includes: docs, code analysis, examples, troubleshooting

Multi-version documentation:

# Track multiple versions
skill-seekers scrape --url "https://docs.react.dev" --versions "17,18,19"
# Generate version-aware skills

Заключение

Skill Seekers революционизирует процесс создания AI skills:

Что делает его unique:

  • Multi-source intelligence** — docs + code + PDFs unified
  • AI-powered enhancement** — не просто scraping, а intelligent processing
  • Quality-driven approach** — assessment и optimization built-in
  • Production-ready output** — ready для Claude Code integration
  • Maintenance automation** — incremental updates, не full rewrites

Real impact:

  • Reduces skill creation time от days до minutes
  • Ensures comprehensive coverage источников
  • Maintains quality standards автоматически
  • Keeps skills updated с source changes

GitHub: https://github.com/yusufkaraaslan/Skill_Seekers

Результат: Any documentation или repository becomes a Claude Code skill in minutes, not days. The democratization of AI skill creation.

*Skill Seekers: когда ИИ учится from любой документации automatically.* 🧪

TG

> Пока нет комментариев