AI & LLMs · Beginner
Agent Skills
A packaged set of instructions for an AI agent: when to use it, what steps to follow, which tools to call. Once loaded, the AI works to that script instead of improvising.
TL;DRIn 30 seconds
- A packaged instruction manual for an AI agent: when to use it, what steps to take, which tools to call.
- It captures the know-how of a particular job, so whoever loads it is up to speed without being taught from scratch.
- It's essentially a document — not a plugin, and not a different model.
OverviewIn detail
Agent Skills are best understood as instruction manuals written for an AI agent. A skill sets out when to use it, what steps to follow and which tools to call. It is not a throwaway remark; it is something organised that you can pass to someone else and reuse.
A clever model on its own is not enough — it does not know the ins and outs of your industry or your team's habits. Capturing that experience in a skill file is like handing the AI an onboarding manual: whoever loads it is up to speed, and you do not have to teach it from scratch every time.
A concrete example: a skill can specify that when a user uploads a CV, the AI scores it line by line against the job requirements using the evaluation method the skill defines, and suggests edits. Loaded with that skill, the AI follows that method rather than offering a few vague compliments. Writing weekly reports, analysing competitors, reviewing contracts — anything whose process you can spell out can be written as a skill too.
A skill is not a plugin and not a model. It is usually just a written document that the AI reads and follows. That makes it light, reusable and shareable — but how well it works depends on how clearly the manual itself is written. Nailing down the steps and the boundaries does more than piling up pleasant-sounding words.
The difference from a prompt is that a prompt is often a one-off instruction, while a skill is more complete — usually with reference material and a fixed process, reusable and shareable with colleagues. A well-written skill gives steady results whoever uses it and whichever model runs it.
MythsCommon misunderstandings
- ✕Thinking installing a skill swaps in a smarter model ✓A skill only makes the same model work to your rules; its abilities are unchanged
- ✕Thinking a skill can inject new knowledge ✓It sets out methods and steps — facts the model doesn't know, it still doesn't know
- ✕Thinking more skills is always better ✓Past a point, deciding which to load and when becomes a new burden of its own
In practiceWhere you will meet it
- Write your team's evaluation standards as a skill so the AI scores every proposal by the same ruler
- Capture the steps of a common task so a different colleague gets the same result
- Load the relevant skill in a client that supports them, then set it to work on that kind of job
SummaryThe key points
- A packaged instruction set for an AI agent
- Sets out when it applies, the steps, and which tools to call
- Once loaded, the AI works to that script
- It's a document at heart — not a plugin or a model
FAQ
What does Agent Skills mean?
A packaged instruction manual for an AI agent: when to use it, what steps to take, which tools to call.It captures the know-how of a particular job, so whoever loads it is up to speed without being taught from scratch.It's essentially a document — not a plugin, and not a different model.
Where will I run into Agent Skills?
Write your team's evaluation standards as a skill so the AI scores every proposal by the same ruler;Capture the steps of a common task so a different colleague gets the same result;Load the relevant skill in a client that supports them, then set it to work on that kind of job。
What do people most often get wrong about Agent Skills?
Thinking installing a skill swaps in a smarter model — but A skill only makes the same model work to your rules; its abilities are unchanged;Thinking a skill can inject new knowledge — but It sets out methods and steps — facts the model doesn't know, it still doesn't know。
What are the key points about Agent Skills?
A packaged instruction set for an AI agent;Sets out when it applies, the steps, and which tools to call;Once loaded, the AI works to that script;It's a document at heart — not a plugin or a model。
Common pitfalls A skill pack is mostly just a written document the AI reads and follows — it isn't a new model.
Entry 11 of 186 · All entries
Keep readingRelated reading
AI Agent
An AI program that sets its own goal, breaks it into steps and calls tools to get the job done. You hand it a task and it works out how to do it — instead of only answering one question at a time.
LLM
A very large program that learned to predict the next piece of text after reading an ocean of it. Whatever you type, it is guessing what is most likely to come next — writing, translating and answering questions are all the same thing used differently.
嵌入(Embedding)
把一段文字换算成一串数字,让电脑能比较哪两句话意思更接近。搜索和知识库能在换了说法时也找对答案,靠的就是它。
Claude
A large language model built by Anthropic, known for understanding long documents, writing well and writing code. It comes both as a chat assistant on the web and phone, and in a dedicated form for programming.
Prompt
The message you send to an AI — the instruction or question you give it. How well that passage is written decides whether the answer that comes back is on target.
MCP
A standard interface that lets AI connect to outside tools and data in one uniform way. Build a tool to it once and any AI assistant can call it, saving you from writing a separate integration for each tool.
Where to go nextExplore more
Back to the skills themselves
Most of the concepts covered here come from skills that install into your AI client with a single command.