Getting started with AI agent skills

You’ve probably used ChatGPT, Claude, Copilot, or another AI assistant. Maybe you’ve asked it to write an email, explain a concept, or help plan a trip. But have you noticed that some AI tools can do things (search the web, read documents, run code) while others can only chat?

That difference comes down to skills.

AI skills in plain language

Think of an AI assistant as a very smart person sitting at a desk. Without any tools, they can think, talk, and reason, but they can’t look anything up, make phone calls, or access your files.

Skills are the tools on that desk. Each one lets the AI do something specific:

A web search skill lets it look up current information
A file reader skill lets it read your documents
A calculator skill lets it do precise math
A calendar skill lets it check your schedule

The more skills an AI has, the more useful it becomes, just like a person with access to the right tools. Knowing which tools are available and choosing the right one for the job is what separates a frustrating AI experience from a productive one.

When we talk about “skills” on this site, we mean something specific: a markdown file that tells an agent what to do, step by step. Some capabilities come built into your AI tool. Others you can write yourself as a simple text file and share with anyone. The practical advice below applies either way, but understanding that skills are files you can create and customize is what makes them powerful.

What this means for you

Understanding skills changes how you use AI tools.

1. Know what your AI can (and can’t) do

Every AI tool has a different set of skills. Before asking for something, it helps to know what’s available:

AI Tool	Typical skills
ChatGPT	Web browsing, code execution, image generation, file reading
Claude	File reading, code analysis, web search (with tools enabled)
Google Gemini	Google Search, Maps, YouTube, Gmail integration
Microsoft Copilot	Office document editing, web search, image generation

2. Ask in terms of actions, not just questions

Instead of:

“What’s the weather like?”

Try:

“Search for the current weather in Minneapolis and the 5-day forecast.”

The second prompt tells the AI to use its search skill rather than guess from its training data.

3. Chain skills together

You get more out of AI when you combine skills in a single request:

“Read the attached PDF report, summarize the key findings in 3 bullet points, then draft an email to my team sharing those findings.”

That request chains reading, analysis, and writing together in one go.

Practical tips for better results

Be specific about what you want done

Vague requests lead to vague results. Compare:

Vague: “Help me with my presentation” Specific: “Review the attached slide deck. For each slide, suggest improvements to the text for clarity and suggest where a visual would help.”

Tell the AI what format you want

Skills often have multiple ways to present results. Guide the output:

“Give me a bullet-point summary” (vs. a paragraph)
“Format this as a table” (vs. a list)
“Keep the email under 100 words” (vs. no length guidance)

Iterate and refine

AI works best as a conversation, not a one-shot request:

Make your initial request
Review the result
Ask for specific changes: “Make the tone more casual” or “Add a section about pricing”
Repeat until you’re satisfied

Know when to verify

AI is useful but not infallible. Always double-check:

Facts and figures, especially dates, statistics, and claims
Calculations, particularly for anything important
Current events, since training data has a cutoff date
Legal or medical information (always consult professionals for big decisions)

Common AI skills and when to use them

Web search

Use when: You need current information, facts you want verified, or recent news Example: “Search for the latest reviews of the iPhone 16”

Document reading

Use when: You have a file you want analyzed, summarized, or transformed Example: “Read this contract and highlight any unusual clauses”

Code execution

Use when: You need precise calculations, data analysis, or charts Example: “Calculate the monthly payment on a $300,000 mortgage at 6.5% over 30 years”

Image generation

Use when: You need visuals for presentations, social media, or creative projects Example: “Create a simple diagram showing the three steps in our onboarding process”

Writing and editing

Use when: You need drafts, rewrites, proofreading, or tone adjustments Example: “Rewrite this paragraph to be more concise and professional”

Beyond built-in skills: installing and creating your own

The skills listed above come built into their respective platforms. But you’re not limited to what ships out of the box. Many AI tools now let you install custom skills, and some let you create your own by writing a simple text file.

A skill file is a markdown document that tells the agent how to perform a specific task. It describes the steps, the checks, and the expected output in plain language. You don’t need to write code. If you can write clear instructions for a coworker, you can write a skill file. Once you drop it into the right folder, your AI agent can use it just like any built-in skill.

For example, a “review my writing” skill might tell the agent to check for clarity, flag jargon, suggest shorter alternatives for long sentences, and format its feedback as a bulleted list. You write it once, and then anyone on your team can use it.

Browse the skills library for examples of real skill files you can install and try right away.

What’s next

If you want to understand what’s happening behind the scenes when agents use these skills, How agents actually work explains the think-act-observe loop and agent memory patterns covers how agents retain context across conversations.

Ready to put these ideas into practice? Explore our other everyday guides:

Writing and Communication for help with emails, documents, and more
Research and Analysis for researching topics and comparing options
Organization and Productivity for managing tasks, schedules, and information