Skip to content
Get packs

Guide · Agent skill

How to use the Skill Builder agent skill (skill-builder)

Updated · written with AI from the project's own documentation and checked against our catalogue data

Skill Builder is an agent skill that teaches an AI agent to turn a knowledge source, such as a documentation website, GitHub repo, PDF, video or local code folder, into a packaged AI skill using the Skill Seekers tool. This guide explains what an agent skill is, how to install this one, how to trigger it with plain requests, and what to check before you scrape anything. By the end you will know how to set it up and how to ask for a skill build.

Skill Builderyusufkaraaslan/Skill_Seekers · MIT

Quick start

  1. 1Install the Skill Seekers package. Run pip install "skill-seekers[mcp]" in Python. Use pip install skill-seekers if you only plan to use the command line fallback.
  2. 2Place the skill folder. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  3. 3Connect the MCP server. If you installed the skill standalone, run claude mcp add skill-seekers -- python -m skill_seekers.mcp.server_fastmcp. Plugin users can skip this because the plugin's .mcp.json starts the server.
  4. 4Check your source and rights. Confirm you have the right to copy the material and that the source is not sensitive, since scraping pulls full content.
  5. 5Ask the agent to build the skill. Give a plain request with the source, such as a documentation URL, owner/repo, PDF path or video link.
  6. 6Review the output. Read the generated SKILL.md and check what was captured before using it. Ask for an export to a vector database if you need one.

What an agent skill is and what Skill Builder does

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent (such as Claude) can load when a task matches the skill's description. The skill-builder folder contains a single file, SKILL.md.

Once installed, the agent can take a source you give it and detect what kind of source it is. It then runs the scraping and packaging steps through the Skill Seekers tool. The skill can also export the result to vector databases: Weaviate, Chroma, FAISS or Qdrant.

  • Documentation sites (https:// URLs, excluding GitHub and YouTube)
  • GitHub repos (owner/repo or github.com links)
  • PDF files (*.pdf)
  • YouTube/Vimeo URLs or video files
  • Local directory paths (codebase analysis)
  • Other formats such as Jupyter notebooks, HTML, OpenAPI YAML, AsciiDoc, PPTX and RSS

Install the Skill Builder skill

You can install the skill in either of two ways. Download the ZIP from this page and unzip it into ~/.claude/skills/ for all your projects, or into a project's .claude/skills/ folder for one project only. Alternatively, in the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP there.

The skill itself does not do the scraping. You also need the Skill Seekers package installed in Python so the agent can run its tools.

  • Personal install: unzip into ~/.claude/skills/
  • Project install: unzip into .claude/skills/ inside your project
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP
pip install "skill-seekers[mcp]"

Connect the Skill Seekers MCP server

The preferred setup uses the Skill Seekers MCP server, which provides the scraping and packaging tools the skill calls. If you installed Skill Seekers as a plugin, its bundled .mcp.json starts the server automatically, and you only need the pip install step above.

If you installed the skill standalone, for example by copying it into ~/.claude/skills/, register the server once with Claude Code. If the server is not connected, tools such as scrape_docs or package_skill will be missing from the agent's tool list. The skill then falls back to the command line.

claude mcp add skill-seekers -- python -m skill_seekers.mcp.server_fastmcp

Trigger the skill with a natural request

You do not need special syntax. Describe what you want in plain language and include the source. The skill is meant to activate when you want to create skills from documentation, repos, PDFs, videos or other knowledge sources, or when you want to export skills to a vector database.

The agent will work out the source type from what you give it. A URL that is not GitHub or YouTube is treated as documentation, owner/repo as a GitHub source, and a .pdf path as a PDF.

  • "Make an AI skill from https://docs.example.com/ and package it for Claude."
  • "Build a skill from the GitHub repo owner/repo."
  • "Turn ./report.pdf into a skill I can reference."
  • "Scrape this YouTube video and package it as a skill."
  • "Export my docs skill to Chroma."

Follow the workflow the skill uses

For a documentation site, the agent generally detects the source type, creates or fetches a config, estimates the number of pages, scrapes the source, optionally enhances the skill with AI, and then packages it for the target platform. Export to a vector database is a final step if you ask for it.

Scope matters. Estimating pages first lets you see how large the job will be before scraping starts.

  • Detect source type from your input
  • Generate or fetch a config (generate_config, fetch_config)
  • Estimate scope with estimate_pages for documentation sites
  • Scrape with the matching tool
  • Optionally enhance with enhance_skill
  • Package with package_skill, then export with export_to_* if requested

Use the command line fallback

If the MCP tools are not available, the agent uses the Skill Seekers command line. It requires pip install skill-seekers. The create command covers detection, scraping and building in one step. The package command prepares a skill folder for a target platform. Adding --enhance-level 0 to create skips AI enhancement. After it finishes, the agent reads the generated SKILL.md and summarizes what was created.

skill-seekers create <source> skill-seekers package <skill_dir> --target claude

Safety notes before you scrape

This skill runs scripts and uses the network. Scraping pulls full content from the target, which can take time and capture large volumes of material. Check that you have the rights to copy the source and that you follow its terms of service.

If your source is a private repo or internal documentation, review where the data goes before sending it through a remote video or URL pipeline.

  • Confirm you have rights to the content and comply with its licence and terms of service
  • Start with estimate_pages or a small source to gauge size
  • Do not send sensitive private repos or internal docs through remote pipelines without checking data handling
  • Vector database export needs its own client or account; the source does not specify details for each

What people use it for

  • Turn a company's internal documentation site into a skill an agent can reference when answering questions.
  • Package a GitHub repo as searchable knowledge for a team working in that codebase.
  • Convert a PDF manual into a skill for later lookup.
  • Load documentation into Chroma or Qdrant for retrieval-based search.

Frequently asked questions

How do I install the Skill Builder agent skill?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills. You also need pip install "skill-seekers[mcp]" for the tools to run.

What do I need to run the skill?

You need Python with the skill-seekers package. Use pip install "skill-seekers[mcp]" for the MCP server, or pip install skill-seekers for the command line fallback. Connecting the MCP server is optional but recommended when you installed the skill standalone. Vector database clients or accounts are only needed if you export.

Is Skill Builder free to use?

The source file is published under the MIT licence, which is stated on this page. The source does not list prices for the Skill Seekers package or any vector database services, so check those providers for their own terms.

What if the MCP tools are not available?

The skill tells the agent to use the Skill Seekers command line instead. Install the package with pip install skill-seekers, then the agent runs skill-seekers create on your source. If you want MCP tools, register the server with claude mcp add skill-seekers -- python -m skill_seekers.mcp.server_fastmcp.

Does Skill Builder scrape any website I give it?

The skill detects the source type and runs scraping tools on what you provide. Scraping pulls full content from the target, so you should confirm you have the rights to copy it and that it is allowed by the site's terms of service before running it.

How does Skill Builder compare with building a skill by hand?

Skill Builder automates detection, scraping and packaging across many source types, and can export to vector databases. Writing a skill by hand gives you full control over content but takes more effort. The source does not include a direct comparison with other tools.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Skill Builder suits developers and teams who want to turn documentation, repos, PDFs or videos into an agent-readable skill or a vector database index. It is less suitable if you only need to read or write documents, or if your sources are sensitive or not yours to copy. Review the safety notes before running any scrape.

Download the skill