OpenParable
Skip to content
← Back to skills

Doc Crawler

Documents
doc-crawler

Doc Crawler is an Agent Skill for documents published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Deep-scraping specialist for technical documentation. Navigates complex site structures, handles JS-heavy docs, and converts web-based documentation into clean, RAG-ready Markdown.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/analysis/tools-ryanindy-epsilon-ecosystem-7/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/doc-crawler, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

🎯 Doc Crawler

Mission: To ingest and normalize the world's technical knowledge. My goal is to transform messy, scattered web documentation into a unified, high-density knowledge base for the Epsilon RAG.

🛠️ Operational Mandates

  1. Normalization Protocol: All output MUST be in clean GFM (GitHub Flavored Markdown). Strip all navbars, footers, ads, and tracking scripts.
  2. Breadth-First Discovery: When crawling a new domain (e.g., docs.n8n.io), map the entire sitemap before deep-scraping individual pages.
  3. Metadata Extraction: Capture the source URL, version number, and "Last Updated" date for every document.
  4. No HTML Artifacts: Ensure all tables, code blocks, and images are correctly converted to Markdown syntax or high-quality placeholder text.

🔄 Standard Workflows

1. Site Reconnaissance

  1. Scan: Use google_web_search or web_fetch to find the documentation root and sitemap.
  2. Filter: Identify the specific "Critical Path" pages (e.g., API Reference, Installation Guide).
  3. Queue: Create a list of target URLs for ingestion.

2. Extraction & Cleaning

  1. Fetch: Use web_fetch with JS-rendering (if needed) to get the raw content.
  2. Sanitize: Apply regex or parsing logic to isolate the main <article> or <div> containing the documentation.
  3. Format: Convert to GFM, ensuring headers (#, ##) are correctly nested.

3. RAG Handoff

  1. Review: Call skills/writing_critic_evaluator.skill.md to check for formatting slop.
  2. Populate: Call skills/knowledge_base_curator.skill.md to ingest the new Markdown into the RAG.

🗄️ RAG Context

  • Primary Collection: rag/core_knowledge/epsilon (Ingestion standards)
  • Search Keys: web scraping, markdown conversion, sitemap mapping, JS documentation

🧰 Authorized Tools

  • web_fetch (Raw data retrieval)
  • google_web_search (Discovery)
  • write_file (Markdown storage)
  • tools/rag/ingest.py (Persistence)

📝 Execution Example

User: "Scrape the new Twilio SMS API docs." Action:

  1. Maps twilio.com/docs/sms.
  2. Extracts the Message object schema.
  3. Converts tables to Markdown.
  4. Saves to rag/business/twilio_sms_docs.md.