HomeSkills
Skills
A small, curated set of Claude Agent Skills you can drop into a Python project and start using right away.
An Agent Skill is a folder with a SKILL.md file inside, instructions and sometimes scripts, that an agent loads only when a task calls for it. Skills sit outside the prompt until they're needed, then take over that piece of work directly. That's worth tracking on PyStack for the same reason a Python library is: it's real, reusable code that changes how a project gets built.
The list here stays short on purpose. These are the official Anthropic Agent Skills, so they share one source repo, and each has to do something concrete in a Python or AI project and install in one command, ready to use as soon as it lands.
Extracts text and tables from PDFs, fills in forms, and merges or splits documents.
Install
git clone https://github.com/anthropics/skills.git && cp -r skills/skills/pdf .claude/skills/pdfWhy we picked it
PDF handling comes up constantly in AI pipelines around finance and operations: invoices, statements, reports. This skill does the fiddly parts, form fields, merged pages, extracted tables, without a separate library call for each one.
Also evaluated
- pypdfA solid Python library for the same job, but it's code you call yourself rather than a skill an agent picks up on its own.
Creates and edits Word documents, including tracked changes, comments, and formatting that survives the round trip.
Install
git clone https://github.com/anthropics/skills.git && cp -r skills/skills/docx .claude/skills/docxWhy we picked it
Most business writing still lands in Word. This skill lets an agent produce or revise an actual .docx file instead of plain text someone has to reformat by hand.
Also evaluated
- python-docxThe underlying library for working with Word files in Python; this skill wraps it into something an agent can use directly.
Reads, edits, and builds Excel workbooks, including formulas, formatting, and charts.
Install
git clone https://github.com/anthropics/skills.git && cp -r skills/skills/xlsx .claude/skills/xlsxWhy we picked it
Spreadsheets are still where a lot of financial and operational data lives. This skill turns an agent's output into a workbook someone can open and use immediately, not a CSV dump.
Also evaluated
- openpyxlThe Python library doing the work underneath; useful directly if you're scripting rather than delegating to an agent.
Walks an agent through scaffolding, wiring, and testing a new MCP server.
Install
git clone https://github.com/anthropics/skills.git && cp -r skills/skills/mcp-builder .claude/skills/mcp-builderWhy we picked it
Building an MCP server by hand means relearning the same boilerplate every time. This skill keeps the process consistent, which starts to matter once a project has more than one server to maintain.
Drives a running web app in a real browser to click through flows and catch UI regressions.
Install
git clone https://github.com/anthropics/skills.git && cp -r skills/skills/webapp-testing .claude/skills/webapp-testingWhy we picked it
Manual QA is where small teams fall behind first. This gives an agent a repeatable way to check that a page still works after a change, without a full test suite to maintain.
Packages a new capability into a SKILL.md folder that follows Anthropic's own skill format.
Install
git clone https://github.com/anthropics/skills.git && cp -r skills/skills/skill-creator .claude/skills/skill-creatorWhy we picked it
It's the fastest way to turn a one-off instruction into something reusable across projects, and it's the same process that produced every other skill on this list.
What I learned
The clearest benefit shows up over a long session: instead of re-explaining a task to an agent every time it comes up, a good skill just runs it the same way twice.
Expect this list to grow slowly. A skill earns a spot here the same way a tool does, by holding up in a real project.