skills.sh: Pre-Built Claude Skills

Building a Claude skill from scratch takes time. You have to figure out what you want the model to do, write instructions that actually produce that behavior, test them against real work, and iterate until the outputs are consistently useful. That process is worth doing, but it is not a fast path to your first working skill.

skills.sh is a library of pre-built Claude skills that shortcuts the starting position. Instead of beginning from a blank document and a general idea, you start with a working skill that someone else has already debugged and refined, apply it to your actual work, and modify it to fit how you specifically operate.

What the library covers

The skills.sh ecosystem is focused on the work that content creators and marketers actually do. The current library spans a few categories that matter for content production:

SEO. Skills for auditing on-page optimization, evaluating content for search intent, analyzing technical SEO issues, and identifying gaps in topical coverage. These are not generic checklists. They are structured workflows that walk the model through a real audit process rather than asking it to produce a list of things to check.

Content strategy. Skills for building content plans around specific audiences, mapping topics to buyer stages, developing pillar and cluster structures, and identifying what a site should be writing about based on search opportunity and what competitors are not covering.

Copywriting. Skills for landing page copy, email sequences, product descriptions, and conversion-focused writing. The copywriting skills encode specific frameworks rather than vague instructions to "write persuasively," so the output follows an actual structure rather than just sounding like marketing.

Marketing psychology. Skills that apply behavioral frameworks to copy and messaging. If you want the model to think through how loss aversion or social proof should shape a specific piece of content, this is the skill category that makes that explicit rather than hoping the model infers it.

Each skill in the library is a Markdown document with clear instructions, usage guidance, and in many cases example inputs and outputs that show what the skill is designed to produce.

How to use a skill from the library

The workflow is the same as any Claude skill. You take the Markdown document from skills.sh, add it to a Claude project as the custom instructions, and start working. The skill runs in the background on every session within that project.

The difference from writing your own from scratch is that you skip the first few iterations of figuring out whether your instructions are producing anything useful. A skill from the library comes with that baseline already established. You are tuning it rather than building it.

Some skills are worth using as-is for the first several sessions before modifying anything. Others will be immediately obvious that they need adjustment for your specific context. An SEO audit skill written for e-commerce sites is going to behave differently on a content-heavy blog, and it will need some modification before the output is calibrated for what you are actually trying to do.

The more interesting part: building your own

The library is the starting point, not the ceiling.

The real value of the skills ecosystem is that every skill is just a document with instructions. There is no proprietary format, no platform lock-in, no black box. You can read every skill, understand exactly what it is doing, modify any part of it, or use it as a template for something entirely different.

The skills.sh library makes the structure visible. Once you have read through five or six skills, you start to see how they are constructed: what kinds of instructions produce consistent behavior, how to specify format and output structure, how to handle conditional logic in natural language. That pattern literacy is what lets you build skills that do not exist yet.

If your workflow involves a task that no library skill covers, you can write it. If a library skill is close but not quite right, you fork it. The library is a learning resource as much as it is a tool resource.

Why this matters for content producers specifically

Content work is full of repeatable tasks that most people redo from scratch every time. An editorial review. A keyword brief. A content outline. A competitor comparison. These are not creative tasks. They are structured analytical tasks that follow a pattern.

A skill for each of those patterns means you are not reinventing the approach every time you need to do one. The model runs the pattern. You review the output and focus your attention on the judgment calls that actually require your expertise. That is less time in setup mode and more time on the work that cannot be automated.