Back in my day we referred to this as &quot;documentation&quot;. It turns out it&#x27;s actually useful for developers too, not just agents.

Pro tip, just add links in code comments&#x2F;readmes with relevant &quot;skills&quot; for the code in question. It works for both humans and agents.

I’ve seen some dev agents do this pretty well.

&gt; But skills are not fundamentally different from *.instruction.md prompt in Copilot or AGENT.md and its variations.One of the best patterns I’ve see is having an &#x2F;ai-notes folder with files like ‘adding-integration-tests.md’ that contain specialized knowledge suitable for specific tasks. These “skills” can then be inserted&#x2F;linked into prompts where I think they are relevant.But these skills can’t be static. For best results, I observe what knowledge would make the AI better at the skill the next time. Sometimes I ask the AI to propose new learnings to add to the relevant skill files, and I adopt the sensical ones while managing length carefully.Skills are a great concept for specialized knowledge, but they really aren’t a groundbreaking idea. It’s just context engineering.

I like the focus on python cli tools, using the standard argparse module, and writing good help and self documentation.You can develop skills incrementally, starting with just one md file describing how to do something, and no code at first.As you run through it for the first several times, testing and debugging it, you accumulate a rich history of prompts, examples, commands, errors, recovery, backing up and branching. But that chat history is ephemeral, so you need to scoop it up and fold it back into the md instructions.While the experience is still fresh in the chat, have it uplift knowledge from the experience into the md instructions, refine the instructions with more details, give concrete examples of input and output, Add more detailed and explicit instructions, handle exceptions and prerequisites, etc.Then after you have a robust reliable set of instructions and examples for solving a problem (with branches and conditionals and loops to handle different conditions, like installing prerequisite tools, or checking and handling different cases), you can have it rewrite the parts that don&#x27;t require &quot;thought&quot; into python, as a self documenting cli tool that an llm, you, and other scripts can call.It&#x27;s great to end up with a tangible well documented cli tool that you can use yourself interactively, and build on top of with other scripts.Often the whole procedure can be rewritten in python, in which case the md instructions only need to tell how to use the python cli tool you&#x27;ve generated, which cli.py --help will fully document.But if it requires a mix of llm decision making or processing plus easily automated deterministic procedures, then the art is in breaking it up into one or more cli tools and file formats, and having the llm orchestrate them.Finally you can take it all the way into one tool, turn it outside in, and have the python cli tool call out to an llm, instead of being called by an llm, so it can run independently outside of cursor or whatever.It&#x27;s a lot like a &quot;just in time&quot; compiler from md instructions to python code.Anyone can write up (and refine) this &quot;Self Optimizing Skills&quot; approach in another md file of meta instructions for incrementally bootstrapping md instructions into python clis.

MCP servers are really just skills paired with python scripts, it&#x27;s not really that different, MCP just lets you package them together for distribution.

The general idea is not very new, but the current chat apps have added features that are big enablers.That is, skills make the most sense when paired with a Python script or cli that the skill uses. Nowadays most of the AI model providers have code execution environments that the models can use.Previously, you could only use such skills with locally running agent clis.This is imo the big enabler, which may totally mean that “skills will go big”. And yeah, having implemented multiple MCP servers, I think skills are a way better approach for most use-cases.

It is not an invention if it is common sense and there is plenty of previous art. How would you otherwise dynamically extend the prompt? You will have some kind of function that based on the selected preferences add more prompt to the base prompt. That is basically what this is except that Anthropic added it as a built in tool.

Many useful inventions seem blindingly obvious in hindsight.Yes, in the end skills are just another way to manage prompts and avoid cluttering the context of a model, but they happen to be one that works really well.

With a little bit of experience, I realized that it makes sense even for agent to run commands&#x2F;scripts for deterministic tasks. For example, to find a particular app out of a list of N (can be 100) with a complex filtering crietria, best option is to run a shell command to get specific output.Like this, you can divide a job to be done into blocks of reasoning and deterministic tasks. The later are scripts&#x2F;commands. The whole package is called skills.

No. You basically call a function to temporarily or permanently extend the base prompt. But of course you can think of other patterns to do more interesting things depending on your use-case. The prompt selection is a RAG.

&gt; [The] way to do is to register a tool &#x2F; function to load and extend the base prompt and presto - you have implemented your own version of skills.So are they basically just function tool calls whose return value is a constant string? Do we know if that’s how they’re implemented, or is the string inserted into the new input context as something other than a function_call_output?

&gt; However, skills are different from MCP. Skills has nothing to do with tool calling at allAlthough skills require that you have certain tools available like basic file system operations so the model can read the skills files. Usually this is implemented as ephemeral &quot;sandbox environment&quot; where LLM have access to file system and can also execute python, run bash commands etc.

added blog post about skills in AI and references toDotprompt &#x2F; Claude &#x2F; Dia browser skills - &quot;Skills Everywhere: Portable Playbooks for Codex, Claude, and Dia&quot;<a href="https:&#x2F;&#x2F;pythonic.ninja&#x2F;blog&#x2F;2025-12-14-codex-skills&#x2F;" rel="nofollow">https:&#x2F;&#x2F;pythonic.ninja&#x2F;blog&#x2F;2025-12-14-codex-skills&#x2F;</a>

Isn&#x27;t the simplicity of the concept, that will make it &quot;explode&quot;?

LLMs need prompts. Prompts can get very big very quickly. The so called &quot;skills&quot;, which exist in other forms in other platforms outside of Anthropic and OpenAI, are simply a mechanism to extend the prompt dynamically. The tool (scripts) that are part of the skill are no different then simply having the tools already installed in the OS where the agent operates.The idea behind skills is sound because context management matters.However, skills are different from MCP. Skills has nothing to do with tool calling at all!You can implement your own version of skills easily and there is absolutely zero need for any kind of standard or a framework of sorts. They way to do is to register a tool &#x2F; function to load and extend the base prompt and presto - you have implemented your own version of skills.In ChatBotKit AI Widget we even have our own version of that for both the server and when building client-side applications.With client-side applications the whole thing is implemented with a simple react hook that adds the necessary tools to extend the prompt dynamically. You can easily come up with your own implementation of that with 20-30 lines of code. It is not complicated.Very often people latch on some idea thinking this is the next big thing hoping that it will explode. It is not new and it wont explode! It is just part of a suite of tools that already exist in various forms. The mechanic is so simple at its core that practically makes no sense to call it a standard and there is absolutely zero need to have it for most types of applications. It does make sense for coding assistant though as they work with quite a bit of data so there it matters. But skills are not fundamentally different from *.instruction.md prompt in Copilot or AGENT.md and its variations.

Skills are lazy loaded prompt engineering. They are simple, but powerful. Claude sees a one line index entry per skill. You can create hundreds. The full instructions only load when invoked.Those instructions can reference external scripts that Claude executes without loading the source. You can package them with hooks and agents in plugins. You pay tokens for the output, not the code that calls it.Install five MCPs and you&#x27;ve burned a large chunk of tokens before typing a prompt. With skills, you only pay for what you use.You can call deterministic code (pipelines, APIs, domain logic) with a non-deterministic model, triggered by plain language, without the context bloat.

Nagel&#x27;s point is that he could not know what it was like to be a bat.

Huh <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;What_Is_It_Like_to_Be_a_Bat%3F" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;What_Is_It_Like_to_Be_a_Bat%3F</a>

It&#x27;s surprising to me that Anthropic&#x27;s CEO is the only one getting real recognition for their advances. The people around him seem to be as or more crucial for their mission.

They are the LLM whisperers.In the same way Nagel knew what it was like to be a bat, Anthropic has the highest fraction of people who approximately know what it&#x27;s like to be a frontier ai model.

Anthropic is in a bit of a rough spot if you look at the raw data points we have available. Their valuation is in the same order of magnitude as OpenAI, but they have orders of magnitude fewer users. And current leaderboards for famous unsolved benchmarks like ARC AGI and HLE are also dominated by Google and OpenAI. Announcements like the one you linked are the only way for Anthropic to stay in the news cycle and justify its valuation to investors. Their IPO rumours are yet another example of this. But I really wonder how long that strategy can keep working.

I just re-binge-watched Silicon Valley in its entirety, with the benefit of a decade of hindsight, so I could get all the interconnected characters and sub-plots and cultural references together in my head better than the first time I watched it in real time at one episode per month.It really should be required viewing for anyone in the industry, it has so much spot-on social commentary, it&#x27;s just not &quot;tecthical&quot; not to be fully aware of it, even if it stings.<a href="https:&#x2F;&#x2F;silicon-valley.fandom.com&#x2F;wiki&#x2F;Tethics" rel="nofollow">https:&#x2F;&#x2F;silicon-valley.fandom.com&#x2F;wiki&#x2F;Tethics</a>&gt;Meanwhile, Gavin Belson (Matt Ross) comes up with a code of ethics for tech, which he lamely calls &quot;tethics&quot;, and urges all tech CEOs to sign a pledge to abide by the tethics code. Richard refuses to sign, he considers the pledge to be unenforceable and meaningless.&gt;Belson invites Richard to the inauguration of the Gavin Belson Institute for Tethics. Before Belson&#x27;s speech, Richard confronts the former Hooli CEO with the fact that the tethics pledge is a stream of brazenly plagiarized banalities, much like Belson&#x27;s novel Cold Ice Cream &amp; Hot Kisses.&gt;Once at the podium, Belson discards his planned speech and instead confesses to his misdeeds when he was CEO of Hooli. Belson urges California&#x27;s attorney general to open an investigation.&gt;Richard mistakenly thinks that Belson is repentant for all his past bad behavior. But, as Ron LaFlamme (Ben Feldman) explains, Belson&#x27;s contrite act is just another effort to sandbag Richard. If the attorney general finds that Belson acted unethically during his tenure as Hooli CEO, the current Hooli CEO would be the one who has to pay the fine. And since Pied Piper absorbed Hooli, it would be Pied Piper that has to pay the fine.

&gt; <a href="https:&#x2F;&#x2F;www.anthropic.com&#x2F;news&#x2F;donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation" rel="nofollow">https:&#x2F;&#x2F;www.anthropic.com&#x2F;news&#x2F;donating-the-model-context-pr...</a>This is a prime example of what you&#x27;re saying. Creating a &quot;foundation&quot; for a protocol created an year ago that&#x27;s not even a protocolHas the Gavin Belson tecthics energy

Skills are not just markdown files. They are markdown files combined with code and data, which only work universally when you have a general purpose cloud-based code execution environment.Out of the box Claude skills can call python scripts that load modules from Pypi or even GitHub, potentially ones that include data like sqlite files or parquet tables.Not just in Claude Code. Anywhere, including the mobile app.

... I am pretty sure that the name &quot;Anthropic&quot; is as in &quot;principle&quot; not as in &quot;pertaining to human beings&quot;.

Their name is Anthropic. Their entire schtick is a weird humanization of AIs.MCP&#x2F;Tool use, Skills, and I&#x27;m sure others that I can&#x27;t think of.This is might be because of some core direction that is more coherent than other labs.

MCP is a terribly designed (and I assume vibe-designed) protocol. Give me the requirements that an LLM needs to be able to load tools dynamically from another server and invoke them like an RPC, and I could give you a much simpler, better solution.The modern HTTP Streamable version is light-years better, but took a year and was championed by outside engineers faced with the real problem of integrating it, and I imagine was designed by a human.OpenAI was there first, but unfortunately the models weren&#x27;t quite good enough yet, so their far superior approach unfortunately didn&#x27;t take off.

Id argue that this isn&#x27;t so much a fault of the MCP spec but how 95% of AI &#x27;engineers&#x27; have no engineering background. MCP is just an OpenAPI spec. It&#x27;s the same as any other API. If you are exposing sensitive data without any authz&#x2F;n that&#x27;s on the developer.

MCP biggest problem is not being simple

Also, MCP is a serious security disaster. Too simple, I&#x27;d wager.

Also `CLAUDE.md` (which is `AGENTS.md` everywhere? else now)

I get the impression the innovation drivers at OpenAI have all moved on and the people that have moved in were the ones chasing the money, the rest is history.

I like this line of analogy. The next obvious step would be IRC (or microservices?) of AI (for co-reasoning) which could offer the space for specialized LLMs rather than the current approach of monoliths.

I noticed something like this earlier, in the android app you can have it rewrite a paragraph, and then and only then do you have the option to send that as a text message. It&#x27;s just a button that pops up. Claude has an elegance to it.

Well, my MCP servers only really started working when I implemented the prompt endpoints, so I’m happy I’ll never have to use MCP again if this sticks.

It’s the only AI company that isn’t monetize at all costs. I’m curious how deep their culture goes as it’s remarkable they even have any discernible value system in today’s business world.

Anthropic are using Ai beyond the chat window. Without external information, context and tools the &quot;magic&quot; of Ai evaporates after a few minutes.

A good example of:Build things and then talk about them in a way that people remember and share it with friends.I guess some call it clever product marketing.

I was very skeptical about anything not OpenAI for a while, and then discovered Claude code, Anthropic blogposts, etc. It&#x27;s basically the coolest company in the field.

Oh yeah I forgot the biggest one. Claude fucking code. Lol

I&#x27;d credit tool calling to the ReAct paper, which was Princeton CA and Google DeepMind: <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2210.03629" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2210.03629</a>

I hate to be that guy but skills are not an invention of sorts. It a simple mechanism that exists already in many places.The biggest unlock was tool calling that was in invented at OpenAI.

Anthropic has good marketing, but ironically their well marketed mediocre ideas retard development of better standards.

It’s crazy how Anthropic keeps coming up with sticky “so simple it seems obvious” product innovations and OpenAI plays catch up. MCP is barely a protocol. Skills are just md files. But they seem to have a knack for framing things in a way that just makes sense.

Well this explains the weird behaviour of GPT-5 often ignoring a large part of my prompt when I attatched many code&#x2F;csv files despite keeping total token count under control. That is with Github Copilot inside VSCode.The fix was to just switch to Claude 3.5 and now to 4.5 in VSCode.

ugh this is so amateurish. i swear since the release of o3 this has been happening on and off.

Are you sure the that “extended thinking” option from the ChatGPT web client is something different?

And the Xhigh version is only available via API, not chatgpt.

&quot;Oh sorry guys, we made the mistake again that saves us X% in compute cost, we will fix it soon!&quot;

A public warning about OpenAI&#x27;s Plus chat subscription as of today.They advertise 196k tokens context length[1], but you can&#x27;t submit more than ~50k tokens in one prompt. If you do, the prompt goes through, but they chop off the right-hand-side of your prompt (something like _tokens[:50000]) before calling the model.This is the same &quot;bug&quot; that existed 4 months ago with GPT-5.0 which they &quot;fixed&quot; only after some high-profile Twitter influencers made noise about it. I haven&#x27;t been a subscriber for a while, but I re-subscribed recently and discovered that the &quot;bug&quot; is back.Anyone with a Plus sub can replicate this by generating &gt; 50k tokens of noise then asking it &quot;what is 1+1?&quot;. It won&#x27;t answer.[1] <a href="https:&#x2F;&#x2F;help.openai.com&#x2F;en&#x2F;articles&#x2F;11909943-gpt-52-in-chatgpt" rel="nofollow">https:&#x2F;&#x2F;help.openai.com&#x2F;en&#x2F;articles&#x2F;11909943-gpt-52-in-chatg...</a>

Awww. If there weren’t only 237 of them, I would want to bring one of them home.&gt; Kākāpō can be up to 64 cm (25 in) long. They have a combination of unique traits among parrots: finely blotched yellow-green plumage, a distinct facial disc, owl-style forward-facing eyes with surrounding discs of specially-textured feathers, a large grey beak, short legs, large blue feet, relatively short wings and a short tail. It is the world&#x27;s only flightless parrot, the world&#x27;s heaviest parrot, and also is nocturnal, herbivorous, visibly sexually dimorphic in body size, has a low basal metabolic rate, and does not have male parental care. It is the only parrot to have a polygynous lek breeding system. It is also possibly one of the world&#x27;s longest-living birds, with a reported lifespan of up to 100 years.<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;K%C4%81k%C4%81p%C5%8D" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;K%C4%81k%C4%81p%C5%8D</a>

and this is my excuse to talk about the :partyparrot: emoji being from an actual real life documentary <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=9T1vfsHYiKY&amp;pp=ygUSa2FrYXBvIG1hdGluZyBoZWFk" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=9T1vfsHYiKY&amp;pp=ygUSa2FrYXBvI...</a>

And so the Kākāpō Benchmark was born

I&#x27;m impressed you have never encountered :partyparrot: in your work Slack.

TIL about a large moss green flightless parrot :)

They already ride British nature photographers—what do they need bikes for?

as an svg you mean? cause nano banana rides circles around the pelicans

Will Kākāpō be riding bicycles soon?

These work well with local LLMs that are powerful enough to run a coding agent environment with a decent amount of context over longer loops.The OpenAI GPT OSS models can drive Codex CLI, so they should be able to do this.I have high hopes for Mistral&#x27;s Devstral 2 but I&#x27;ve not run that locally yet.

&gt; Skills are a keeper #Good thinking, I agree actually, however..&gt; Skills are based on a very light specification, if you could even call it that, but I still think it would be good for these to be formally documented somewhere.Like a lot of posts around AI, and I hope OP can speak to it, surely you can agree that while when used for a good cool idea, it can also be used for the inverse and probably to more detrimental reason. Why would they document an unmanageable feature that may be consumed.Shareholder value might not go up if they learnt that the major product is learning bad things.Have you or would you try this on a local LLM instead ?

I had a bunch of fun writing about this one, mainly because it was a great excuse to highlight the excellent news about Kākāpō breeding season this year.(I&#x27;m not just about pelicans.)

Apologies for not reading all of your blogs on this, but a follow-up question. Are models still prone to reading these and disregarding them even if they should be used for a task?Reason I ask is because a while back I had similar sections in my CLAUDE.md and it would either acknowledge and not use or just ignore them sometimes. I&#x27;m assuming that&#x27;s more of an issue of too much context and now skill-level files like this will reduce that effect?

why did this simple idea take so long to become available? I remember even in llama 2 days I was doing this stuff, and that model didn&#x27;t even function call.

&gt; On startup Claude Code &#x2F; Codex CLI etc scan all available skills folders and extract just those descriptions into the context. Then, if you ask them to do something that&#x27;s covered by a skill, they read the rest of that markdown file on demand before going ahead with the task.Maybe I still don&#x27;t understand the mechanics - this happens &quot;on startup&quot;, every time a new conversation starts? Models go through the trouble of doing ls&#x2F;cat&#x2F;extraction of descriptions to bring into context? If so it&#x27;s happening lightning fast and I somehow don&#x27;t notice.Why not just include those descriptions within some level of system prompt?

Do skills get access to the current context or are they a blank slate?

Have you used AWS bedrock? I assume these get pretty affordable with prompt caching...

It&#x27;s a folder with a markdown file in it plus optional additional reference files and executable scripts.The clever part is that the markdown file has a section in it like this:
<a href="https:&#x2F;&#x2F;github.com&#x2F;datasette&#x2F;skill&#x2F;blob&#x2F;a63d8a2ddac9db8225ee3b2b62724932f6ca49c6&#x2F;SKILL.md?plain=1#L1-L4" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;datasette&#x2F;skill&#x2F;blob&#x2F;a63d8a2ddac9db8225ee...</a><pre><code> ---
 name: datasette-plugins
 description: &quot;Writing Datasette plugins using Python and the pluggy plugin system. Use when Claude needs to: (1) Create a new Datasette plugin, (2) Implement plugin hooks like prepare_connection, register_routes, render_cell, etc., (3) Add custom SQL functions, (4) Create custom output renderers, (5) Add authentication or permissions logic, (6) Extend Datasette&#x27;s UI with menus, actions, or templates, (7) Package a plugin for distribution on PyPI&quot;
 ---
</code></pre>
On startup Claude Code &#x2F; Codex CLI etc scan all available skills folders and extract just those descriptions into the context. Then, if you ask them to do something that&#x27;s covered by a skill, they read the rest of that markdown file on demand before going ahead with the task.

How do you manage the credentials to requests your CI server in this case?
They are hardcoded in the script associated to your SKILL?

Skills have a lot of uses, but one in particular I like is replacing one off MCP server usage. You can use (or write) an MCP server for you CI system and then add the instructions to your AGENTS.md to query the CI MCP for build results for the current branch. Then you need to find a way to distribute the MCP server so the rest of the team can use it or cook it into your dev environment setup. But all you really care about is one tool in the MCP server, the build result. Or...You can hack together a shell, python, whatever script that fetches build results from your CI server, dumps them to stdout in a semi structured format like markdown, then add a 10-15 line SKILL.md and you have the same functionality -- the skill just executes the one-off script and reads the output. You package the skill with the script, usually in a directory in the project you are working on, but you can also distribute them as plugins (bundles) that claud code can install from a &quot;repository&quot;, which can just be a private git repo.It&#x27;s a little UNIX-y in a way, little tools that pipe output to another tool and they are useful in a standalone context or in a chain of tools. Whereas MCP is a full blown RPC environment (that has it&#x27;s uses, where appropriate).

I&#x27;m so excited for the future, because _clearly_ our technology has loads to improve. Even if new models don&#x27;t come out, the tooling we build upon them, and the way we use them, is sure to improve.One particular way I can imagine this is with some sort of &quot;multipass makeshift attention system&quot; built on top of the mechanisms we have today. I think for sure we can store the available skills in one place and look only at the last part of the query, asking the model the question: &quot;Given this small, self-contained bit of the conversation, do you think any of these skills is a prime candidate to be used?&quot; or &quot;Do you need a little bit more context to make that decision?&quot;. We then pass along that model&#x27;s final answer as a suggestion to the actual model creating the answer. There is a delicate balance between &quot;leading the model on&quot; with imperfect information (because we cut the context), and actually &quot;focusing it&quot; on the task at hand, and the skill selection&quot;. Well, and, of course, there&#x27;s the issue of time and cost.I actually believe we will see several solutions make use of techniques such as this, where some model determines what the &quot;big context&quot; model should be focusing on as part of its larger context (in which it may get lost).In many ways, this is similar to what modern agents already do. cursor doesn&#x27;t keep files in the context: it constantly re-reads only the parts it believes are important. But I think it might be useful to keep the files in the context (so we don&#x27;t make an egregious mistake) at the same time that we also find what parts of the context are more important and re-feed them to the model or highlight them somehow.

I&#x27;m kinda confused about why this even is something that we need an extra feature for when it&#x27;s basically already built in to the agentic development feature. I just keep a folder of md files and I add whatever one is relevant when it&#x27;s relevant. It&#x27;s kinda straight forward to do...Just like you I don&#x27;t edit much in these files on my own. Mostly just ask the model to update an md file whenever I think we&#x27;ve figured out something new, so the learning sticks. I have files for test writing, backend route writing, db migration writing, frontend component writing etc. Whenever a section gets too big to live in agents.md it gets it&#x27;s own file.

Claude Code is not very good at “remembering” its skills.Maybe they get compacted out of the context.But you can call upon them manually. I often do something like “using your Image Manipulation skill, make the icons from image.png”Or “use your web design skill to create a design for the front end”Tbh i do like that.I also get Claude to write its own skills. “Using what we learned about from this task, write a skill document called &#x2F;whatever&#x2F;using your writing skills skill”I have a GitHub template including my skills and commands, if you want to see them.<a href="https:&#x2F;&#x2F;github.com&#x2F;lawless-m&#x2F;claude-skills" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;lawless-m&#x2F;claude-skills</a>

&gt; I&#x27;m a bit of a control freak so I&#x27;ll usually explicitly direct claude to &quot;load the wireframe-skill&quot; and then do X.You shouldn&#x27;t do this, it&#x27;s generally considered bad practice.You should be optimizing your skill description. Often times if I am working with Claude Code and it doesn&#x27;t load I skill, I ask it why it missed the skill. It will guide me to improving the skill description so that it is picked up properly next time.This iteration on skill description has allowed skills to stay out of context until they are needed rather predictably for me so far.

My understanding is that use of &quot;description&quot; frontmatter is essential, bc Claude Code can read just the description without loading the entire file into context.

My understanding is this: A skill is made up of SKILL.md which is what tells claude how and when to use this skill. I&#x27;m a bit of a control freak so I&#x27;ll usually explicitly direct claude to &quot;load the wireframe-skill&quot; and then do X.Now SKILL.md can have references to more finegrained behaviors or capabilities of our skill. My skills generally tend to have a reference&#x2F;{workflows,tools,standards,testing-guide,routing,api-integration}.md. These references are what then gets &quot;progressively loaded&quot; into the context.Say I asked claude to use the wireframe-skill to create profileView mockup. While creating the wireframe, claude will need to figure out what API endpoints are available&#x2F;relevant for the profileView and the response types etc. It&#x27;s at this point that claude reads the references&#x2F;api-integration.md file from the wireframe skill.After a while I found I didn&#x27;t like the progressive loading so I usually direct claude to load all references in the skill before proceeding - this usually takes up maybe 20k to 30k tokens, but the accuracy and precision (imagined or otherwise ha!) is worth it for my use cases.

Easy, let me try to explain:
You want to achieve X, so you ask your AI companion, &quot;How do I do X?&quot;
Your companion thinks and tries a couple of things, and they eventually work.
So you say, &quot;You know what, next time, instead of figuring it out, just do this&quot;... that is a skill. A recipe for how to do things.

Yes. I find these very useful for enforcing e.g. skills like debugging, committing code, make prs, responding to pr feedback from ai review agents, etc. without constantly polluting the context window.So when it&#x27;s time to commit, make sure you run these checks, write a good commit message, etc.Debugging is especially useful since AI agents can often go off the rails and go into loops rewriting code - so it&#x27;s in a skill I can push for &quot;read the log messages. Inserting some more useful debug assertions to isolate the failure. Write some more unit tests that are more specific.&quot; Etc.

I&#x27;m excited to use this with the Ghidra cli mode to rapidly decompile physics engines from various games. Do I want my flight simulator to behave like the Cessna like in flight simulator 3.0 in the air? Codex can already do that. Do I want the plane to handle like Yoshi from Mario Kart 64 when taxiing? It hasn&#x27;t been done yet but Claude code is apparently pretty good at pulling apart n64 roms so that seems within the realm of possibility.

I think it’s also important to think of skills in the context of tasks, so when you want an agent to perform a specialized task, then this is the context, the resources and scripts it needs to perform the task.

I&#x27;m not sure if I have the right mental model for a &quot;skill&quot;. It&#x27;s basically a context-management tool? Like a skill is a brief description of something, and if the model decides it wants the skill based on that description, then it pulls in the rest of whatever amorphous stuff the skill has, scripts, documents, what have you. Is this the right way to think about it?

Yeah, there are a whole lot of AI wrapper applications that could be a folder with a markdown file in at this point!

From a purely technical view, skills are just an automated way to introduce user and system prompt stuffing into the context right? Not to belittle this, but rather that seems like a way of reducing the need for AI wrapper apps since most AI wrappers just do systematic user and system prompt stuffing + potentially RAG + potentially MCP.

When there is no alt text do you have like a solution for that? Like VLMs are really powerful, I imagine they can be used to parse through the unlabeled images automatically if needed.

@simonw Thank you for always setting alt text in your images. I really appreciate it.

If anyone wants to use skills with any other model or tool like Gemini CLI etc. I had created open-skills, which lets you use skills for any other llm.Caveat: needs mac to runBonus: it runs it locally in a container, not on cloud nor directly on mac1. Open-Skills: <a href="https:&#x2F;&#x2F;GitHub.com&#x2F;BandarLabs&#x2F;open-skills" rel="nofollow">https:&#x2F;&#x2F;GitHub.com&#x2F;BandarLabs&#x2F;open-skills</a>

we just released Anthropic&#x27;s Skills talk for those who want to find more info on the design thinking &#x2F; capabilities: <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=CEvIs9y1uog&amp;t=2s" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=CEvIs9y1uog&amp;t=2s</a>

The example Datasette plugin authoring skill I used in my article was entirely written by Claude Opus 4.5 - I uploaded a zip file to its the Datasette repo in it (after it failed to clone that itself for some weird environment reason) and had it use its skill-writing skill to create the rest: <a href="https:&#x2F;&#x2F;claude.ai&#x2F;share&#x2F;0a9b369b-f868-4065-91d1-fd646c5db3f4" rel="nofollow">https:&#x2F;&#x2F;claude.ai&#x2F;share&#x2F;0a9b369b-f868-4065-91d1-fd646c5db3f4</a>

Completely agree with both points. Skills replacing one-off microservices and agents writing their own skills feel like two sides of the same coin to me.
I’m a solo developer building a markdown-first slide editing app. The core format is just Markdown with --- slide separators, but it has custom HTML comment directives for layouts (&lt;!-- layout: title --&gt;, &lt;!-- layout: split --&gt;, etc.) and content-type detection for tables, code blocks, and Mermaid diagrams. It’s a small DSL, but enough that an LLM without context will generate slides that don’t render optimally.
Right now my app is designed for copy-paste from external LLMs, which means users have to manually include the format spec in their prompts every time. But your comment about agents writing skills made me realize the better path: I could just ask Claude Code to read my parser and layout components, then generate a Slide_Syntax_Guide skill for me. The agent already understands the codebase—it can write the definitive spec better than I could document it manually.

Agree. I&#x27;d add that a aha moment to skills is AI agents are pretty good at writing skills. Let&#x27;s say you have developed an involved prompt that explains how to hit an API (possibly with the complexity of reading credentials from an env var or config file) or run a tool locally to get some output you want the agent to analyze (example, downloading two versions of python packages and diffing them to analyze changes). Usually the agent reading the prompt it&#x27;s going to leverage local tools to do it (curl, shell + stdout, git, whatever) every single time. Every time you execute that prompt there is a lot thinking spent on deciding to run these commands and you are burning tokens (and time!). As an eng you know that this is a relatively consistent and deterministic process to fetch the data. And if you were consuming it yourself, you&#x27;d write a script to automate it.So you read about skills (prompt + scripts) to make this more repeatable and reduce time spent thinking. At that point there are two paths you can go down -- write the skill and prompt yourself for the agent to execute -- or better -- just tell the agent to write the skill and prompt and then you lightly edit it and commit it.This may seem obvious to some, but I&#x27;ve seen engineers create skills from scratch because they have a mental model around skills being something that people must build for the agent, whereas IMO skills are you just bridging a productivity gap that the agent can&#x27;t figure out itself (for now), which is instructing it to write tools to automate its own day to day tedium.

CLIs are really good when you can use them. self-documenting, agents already have shell tools, they tend to solve fine-grained auth, etc.feels like the right layer of abstraction for remote APIs

That is effectively what this proposal is about: <a href="https:&#x2F;&#x2F;www.anthropic.com&#x2F;engineering&#x2F;code-execution-with-mcp" rel="nofollow">https:&#x2F;&#x2F;www.anthropic.com&#x2F;engineering&#x2F;code-execution-with-mc...</a>

If only there was a way to progressively disclose the API in MCP instead of presenting the full laundry list up front.

I think the future is likely one that mixes the kitchen-sink style MCP resources with custom skills.Services can provide an MCP-like layer that provides semantic definitions of everything you can do with said service (API + docs).Skills can then be built that combine some subset of the 3rd party interfaces, some bespoke code, etc. and then surface these more context-focused skills to the LLM&#x2F;agent.Couldn’t we just use APIs?Yes, but not every API is documented in the same way. An “MCP-like” registry might be the right abstraction for 3rd parties to expose their services in a semantic-first way.

Can skills completely replace MCPs? For example, can a skill be configured to launch my local Python program in its own venv. I don’t want Claude to spend time spinning up a runtime

MCPs feel complicated. Skills seem to me like the simplest possible design for a mechanism for adding extra capabilities to an existing coding agent.

If we aren’t in the take off phase, I don’t know where we are

I tell Claude to make its own skills. “Which part of this task is worth making a skill for, use your skill making skill to do it”

Skills are just pointers to context so you don&#x27;t need to load all of them upfront, it is as simple as that. By the way cursor rules is effectively the same as agents.md.

This is killing me with complexity.
We had agents.md and were supposed to augment the context there.
Now back to cursor rules and another md file to ingest.

This is great that Codex CLI is adding skills - but it would be far more useful if the CLI looked first in the project (the directory where I&#x27;ve launched codex) `.codex&#x2F;skills` directory and THEN the home directory .codex dir. The same issue exists for prompts.

This is nice, but that it goes into its vendor specific .codex&#x2F; folder is a bit of a drag.I hope such things will be standardized across vendors. Now that they founded the Agentic AI Foundation (AAIF) and also contributed AGENTS.md, I would hope that skills become a logical extension of that.<a href="https:&#x2F;&#x2F;www.linuxfoundation.org&#x2F;press&#x2F;linux-foundation-announces-the-formation-of-the-agentic-ai-foundation" rel="nofollow">https:&#x2F;&#x2F;www.linuxfoundation.org&#x2F;press&#x2F;linux-foundation-annou...</a><a href="https:&#x2F;&#x2F;aaif.io&#x2F;" rel="nofollow">https:&#x2F;&#x2F;aaif.io&#x2F;</a>

I shipped a new release yesterday - <a href="https:&#x2F;&#x2F;llm.datasette.io&#x2F;en&#x2F;stable&#x2F;changelog.html#v0-28" rel="nofollow">https:&#x2F;&#x2F;llm.datasette.io&#x2F;en&#x2F;stable&#x2F;changelog.html#v0-28</a> but yeah, the last core release before that was in August. I&#x27;ve been pushing out plugin releases for it though, for new models from Gemini and Anthropic and others.Honestly the main problem has been that LLM&#x27;s unique selling point back in 2024 was that it was the only tool taking CLI access to LLMs seriously. In 2025 Claude Code and Codex CLI etc all came along and suddenly there&#x27;s not much unique about having a CLI tool for LLMs any more!There&#x27;s also a major redesign needed to the database storage and model abstraction layer in order to handle reasoning traces and more complex tool call patterns. I opened an issue about that here - it&#x27;s something I&#x27;m stewing on but will take quite some work to get right: <a href="https:&#x2F;&#x2F;github.com&#x2F;simonw&#x2F;llm&#x2F;issues&#x2F;1314" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;simonw&#x2F;llm&#x2F;issues&#x2F;1314</a>I&#x27;ve been spending more of my time focusing on other projects which make use of LLM, in particular Datasette plugins that use the asyncio Python library: <a href="https:&#x2F;&#x2F;llm.datasette.io&#x2F;en&#x2F;stable&#x2F;python-api.html#async-models" rel="nofollow">https:&#x2F;&#x2F;llm.datasette.io&#x2F;en&#x2F;stable&#x2F;python-api.html#async-mod...</a>I expect those to drive some core improvements pretty soon.

@simonw: Any particular reason you stopped releasing &quot;llm&quot; regularily? I believe the last release was done in summer. Neither gpt-5.1 nor gpt-5.2 have been added. Are you about to give up on that project? Is it time to search for another one?I also have an open issue since months, which someone wrote a PR for (thanks&quot;) a few weeks ago.Are you still comitted to that project?

I think a tool call can be thought of as special type of reply where it’s contents are parsed and an actual function is called. A skill is more of a dynamic context-enrichment.

Is there a fundamental difference between a skill and a tool or could I just make a terse skill and have that be used in the same way as a tool?

This is a clever abstraction. Reminds me of how tool_use worked in earlier Claude versions - defining a schema upfront and letting the model decide when to call it.Has anyone tested how well this works with code generation in Codex CLI specifically? The latency on skill registration could matter in a typical dev workflow.

this makes sense and as usual great eye by Simon. progressive disclosure is such a powerful concept, that every other tool is going to start adopting it. we&#x27;ve personally had such great results adopting it, especially for performing high quality large migrations.I wrote about this but I&#x27;m certain that eventually commands, MCPs etc will fade out when skills is understood and picked up by everyone<a href="https:&#x2F;&#x2F;kau.sh&#x2F;blog&#x2F;claude-skills&#x2F;" rel="nofollow">https:&#x2F;&#x2F;kau.sh&#x2F;blog&#x2F;claude-skills&#x2F;</a>

It is interesting that they are relying on visual reading for document ingestion instead of OCT. Recently I read an article which says Handwriting recognition has matured, and I&#x27;m beginning to think this is the approach they are takingwirh HAndwiting recognition.

I like Anthropic’s plugin system. I wish everyone would standardize on it instead of everyone having a billion different ways to do slash commmands, skills, etc.

It’s definitely a useful word! Modern tech rollouts often do happen without fanfare. And the word alerts readers to a kind of story shape. So I can see why people use it! Its usage reminds me of when competitive video games develop a new meta, a new powerful technique. There follows a short period where everyone spams the technique over and over. Eventually people figure out a counter and the meta quietly disappears. (Couldn’t help myself!)

Huh! I had not noticed that trend at all.Here&#x27;s the Google Maps article: <a href="https:&#x2F;&#x2F;laurenleek.substack.com&#x2F;p&#x2F;how-google-maps-quietly-allocates" rel="nofollow">https:&#x2F;&#x2F;laurenleek.substack.com&#x2F;p&#x2F;how-google-maps-quietly-al...</a> - note that the Hacker News title left that word out: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46203343">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46203343</a>It&#x27;s possible I was subconsciously influenced by that article (I saw it linked from a few places yesterday I think), but in this case I really did want to emphasize that OpenAI have started doing this without making any announcements about it at all, which I think is noteworthy in its own right.(I&#x27;m also quite enjoying that this may be the second time I&#x27;ve leaked the existence of skills from a major provider - I wrote about Anthropic&#x27;s skills implementation a week before they formally announced it: <a href="https:&#x2F;&#x2F;simonwillison.net&#x2F;2025&#x2F;Oct&#x2F;10&#x2F;claude-skills&#x2F;" rel="nofollow">https:&#x2F;&#x2F;simonwillison.net&#x2F;2025&#x2F;Oct&#x2F;10&#x2F;claude-skills&#x2F;</a>)

Probably in part, yes! Since it’s a pet peeve, I must be making extra special note of it every time I see it. So it’s probable that I’m overweighting it!That being said I’m quite sure that it’s being used more frequently recently. For example, I read a shortish 2000-word article yesterday that uses the word “quietly” four times. And ChatGPT 5.1 used it in most of its responses. Also I’d expect that the frequency illusion wears off quite quickly, whereas I’ve noticed “quietly” for some time and the feeling doesn’t seem to be wearing off. Maybe you’ll start to notice it now too!

Maybe you fell into <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Frequency_illusion" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Frequency_illusion</a> ? :D

Totally unrelated but what’s up with the word “quietly”? Its usage seems to have gone up 5000%, essentially overnight, as if there’s a contagion. You see the word in the New York Times, in government press releases, in blogs. ChatGPT 5.1 itself used the word in almost every single response, and no amount of custom instructions could get it to stop. That “Google Maps of London restaurants” article that’s going around not only uses the word in the headline, but also twice in the closing passage alone, for example. And now Simon, who’s an excellent writer with an assertive style, has started using it in his headlines. What’s the deal? Why have so many excellent writers from a wide range of subjects suddenly all adopted the same verbal tic? Are these writers even aware that they’re doing it?

The skills approach is great for agents and LLMs but I feel agents have to become wider in the context they keep and more proactive in the orchestration.I have been running Claude Code with simple prompts (eg 1) to orchestrate opencode when I do large refactors. I have also tried generating orchestration scripts instead. Like, generate a list of tasks at a high level. Have a script go task by task, create a small task level prompt (use a good model) and pass on the task to agent (with cheaper model). Keeping context low and focused has many benefits. You can use cheaper models for simple, small and well-scoped tasks.This brings me to skills. In my product, nocodo, I am building a heavier agent which will keep track of a project, past prompts, skills needed and use the right agents for the job. Agents are basically a mix of system prompt and tools. All selected on the fly. User does not even have to generate&#x2F;maintain skills docs. I can get them generated and maintained with high quality models from existing code in the project or tasks at hand.1 Example prompt I recently used:
 Please read GitHub issue #9. We have phases clearly marked. Analyze the work and codebase. Use opencode, which is a coding agent installed. Check `opencode --help` about how to run a prompt in non-interactive mode. Pass each phase to opencode, one phase at a time. Add extra context you think is needed to get the work done. Wait for opencode to finish, then review the work for the phase. Do not work on the files directly, use opencodeMy product, nocodo: <a href="https:&#x2F;&#x2F;github.com&#x2F;brainless&#x2F;nocodo" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;brainless&#x2F;nocodo</a>

Visit this link and click on the &quot;Thought for 11m38s&quot; text: <a href="https:&#x2F;&#x2F;chatgpt.com&#x2F;share&#x2F;693ca54b-f770-8006-904b-9f31a585180a" rel="nofollow">https:&#x2F;&#x2F;chatgpt.com&#x2F;share&#x2F;693ca54b-f770-8006-904b-9f31a58518...</a> - that will show you exactly what it spent those 11 minutes doing, most of which was executing Python code using the reportlab library to generate PDF files, then visually inspecting those PDF files and deciding to make further tweaks to the code that generates them.

&gt; It took just over eleven minutes to produce this PDF,Incredibly dumb question, but when they say this, what actually happens?Is it using TeX? Is it producing output using the PDF file spec? Is there some print driver it&#x27;s wired into?

In these new world order, frameworks need to stop changing their APIs for minimal marginal improvements of syntax.

Yes, definitely. I&#x27;ve had a lot of success already showing LLMs short examples of coding libraries they don&#x27;t know about from their core training data.

Does this mean I can point to a code snippet and a link to the related documentation and the coding agent refer to it instead of writing &quot;outdated&quot; code?Some frameworks&#x2F;languages move really fast unfortunately.

Can or should skills be used for managing the documentation of dependencies in a project and the expertise in them?I’ve been playing with doing this but kind of doesn’t feel the most natural fit.

Apparently this is a British vs American English thing. I&#x27;ve decided to stay stubbornly British on this one.

I thought this was the same in all languages (my reference was German) because names are singular terms, even in British English, but apparently there are special rules.

Collective nouns!They vary between British and American English. In this case, either would acceptable depending on your dialect.Also very noticeable with sports teams.American: “Team Spain is going to the final.”British: “Team Spain are going to the final.”<a href="https:&#x2F;&#x2F;editorsmanual.com&#x2F;articles&#x2F;collective-nouns-singular-or-plural&#x2F;" rel="nofollow">https:&#x2F;&#x2F;editorsmanual.com&#x2F;articles&#x2F;collective-nouns-singular...</a>

(Minor grammar note: &quot;OpenAI are&quot; -- it should say &quot;OpenAI is&quot; -- because &quot;OpenAI&quot; is a name and therefore singular.)

I remember you did some reverse engineering when they released agent, does it not feel quite similar to you?I know they didn’t dynamically scan for new skill folders but they did have mentions of the existing folders (slides, docs, …) in the system prompt

Not quite the same thing. Implementing skills specifically means that you have code which, on session start, scans the skills&#x2F;*&#x2F;skill.md files and reads in their description: metadata and loads that into the system prompt, along with an instruction that says &quot;if the user asks about any of these particular things go and read the skills.md file for further instructions&quot;.Here&#x27;s the prompt within Codex CLI that does that: <a href="https:&#x2F;&#x2F;github.com&#x2F;openai&#x2F;codex&#x2F;blob&#x2F;ad7b9d63c326d5c92049abd16f9f5fb64a573a69&#x2F;codex-rs&#x2F;core&#x2F;src&#x2F;skills&#x2F;render.rs#L20-L38" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;openai&#x2F;codex&#x2F;blob&#x2F;ad7b9d63c326d5c92049abd...</a>I extracted that into a Gist to make it easier to read: <a href="https:&#x2F;&#x2F;gist.github.com&#x2F;simonw&#x2F;25f2c3a9e350274bc2b76a79bc8ae8b2" rel="nofollow">https:&#x2F;&#x2F;gist.github.com&#x2F;simonw&#x2F;25f2c3a9e350274bc2b76a79bc8ae...</a>

Hasn’t ChatGPT been supporting skills with a different name for several months now through “agent”?They gave it back then folders with instructions and executable files iirc

Do you have experience using Ghidra in such a way? I’m curious how well it actually performs on that use case.

Giving the llm access to Ghidra so it can directly read and iterate through the Sudoku puzzle that is decompile binaries seems like a good one. Ghidra has a cli mode and various bindings so you can automate decompiling various binaries. For example right now if you want to isolate the physics step of Microsoft flight simulator 3.0 codex will hold your hand and walk you through (over the course of 3-4 hours, using the gui) finding the main loop and making educated guesses about which decompiled c functions in there are likely physics related, but it would be a lot easier to just give it the &quot;Ghidra&quot; skill and say, &quot;isolate the physics engine and export it as a portable cargo package in rust&quot;. If you&#x27;re an NSA analyst you can probably use it to disassemble and isolate interesting behavior of various binaries from state actors a lot faster.

I don&#x27;t know about life-changing but to me there are two major benefits that get me really interested:- Augmenting CLI with specific knowledge and processes: I love the ability to work on my files, but I can only call a smart generalist to do the work. With skills if I want, say, a design review, I can write the process, what I&#x27;m looking for, and design principles I want to highlight rather than the average of every blog post about UX. I created custom gems&#x2F;projects before (with PDFs of all my notes), but I couldn&#x27;t replicate that on CLIs.- Great way to build your library of prompts and build on it: In my org everyone is experimenting with AI but it&#x27;s hard to document and share good processes and tools. With this, the copywriters can work on a &quot;tone of voice&quot; skill, the UX writers can extend it with an &quot;Interface microcopy&quot; skill, and I can add both to my &quot;design review&quot; agent.

Small use case but I’m using skills for analysing and scoring content then producing charts. LLM does the scoring then calls a Python script bundled in the skill that makes a variety of PNG charts based on metrics passed in via command line arguments. Claude presents the generated files for download. The skill.md file explains how to run the analysis and how to call the script and with what options. That way, you can get very consistent charts because they’re generated programmatically, but you can use the LLM for what it’s good at.

The best examples I&#x27;ve seen are still the ones built into ChatGPT and Claude to improve their abilities to edit documents.The Claude frontend-design skill seems pretty good too for getting better HTML+CSS: <a href="https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;skills&#x2F;blob&#x2F;main&#x2F;skills&#x2F;frontend-design&#x2F;SKILL.md" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;skills&#x2F;blob&#x2F;main&#x2F;skills&#x2F;fronte...</a>

I have made a skill that uses Playwright to control Chrome together with functionality to extract HTML, execute JS, click things and most importantly log full network requests. It&#x27;s a blessing for reverse-engineering and making userscripts.

Does anybody have examples of life-changing skills? I can&#x27;t quite understand how they&#x27;re useful, yet...

Something important to keep in mind is the way skills work shouldn&#x27;t be assumed to be the same and work in the same way.

Curious if anyone has applied this &quot;Skills&quot; mindset to how you build your tool calls for your LLM agents applications?Say I have a CMS (I use a thin layer of Vercel AI SDK) and I want to let users interact with it via chat: tag a blog, add an entry, etc, should they be organized into discrete skill units like that? And how do we go about adding progressive discovery?

this is really bothering me that we are building abstractions in our language to hide the fact that most of the features are prompts hardcoded in a text file.

tbh everything about the current implementation of &quot;AI&quot; is starting to look like hot porridge when it comes to real world products.Is the prompting workflow so convenient that it’s worth having to spend twice or thrice as much time double checking the accuracy of the inference and fixing bugs?How long until we collectively decide that to reduce the probability of errors we’re better off going back to writing our own functions, methods, classes etc. because it gives us granular control?Last but not least, we’re devolving to mainframe and terminals…

No I don&#x27;t believe so. Cursor is usually pretty behind other agentic coding tools in my experience.

Oooooo, okay. So in fact it has technical capabilities of utilizing and taking advantage of this information provided as skills. That is much clearer now, I appreciate very much your response.

ChatGPT has had a full Linux container system available to it for nearly three years now.OpenAI keep changing their mind on what to call it. I like the original name, &quot;ChatGPT Code Interpreter&quot;, but they&#x27;ve also called it &quot;advanced data analysis&quot; at various points.Claude added the same feature in September this year: <a href="https:&#x2F;&#x2F;simonwillison.net&#x2F;2025&#x2F;Sep&#x2F;9&#x2F;claude-code-interpreter&#x2F;" rel="nofollow">https:&#x2F;&#x2F;simonwillison.net&#x2F;2025&#x2F;Sep&#x2F;9&#x2F;claude-code-interpreter...</a>In both ChatGPT and Claude you can say things like &quot;use your Python tool to calculate total mortgage payments over a 30 year period for X and Y&quot; and it will write and execute code to do so - but you can also upload files (including CSVs or even SQLite database files) into that container file system and have them write and execute python code to process those in different ways.Skills are just folders full of markdown files that are saved in that container when it first boots up.

So chatgpt can read markdown files?
I am very confused

Yes, provided they are good enough at long chain tool calling to run a coding agent environment and have a decent context length.

Is possible using skills in ofline models?

VS Code Copilot just announced experimental skill support in their November release: <a href="https:&#x2F;&#x2F;code.visualstudio.com&#x2F;updates&#x2F;v1_107#_reuse-your-claude-skills-experimental" rel="nofollow">https:&#x2F;&#x2F;code.visualstudio.com&#x2F;updates&#x2F;v1_107#_reuse-your-cla...</a>

Looks like they added it to the Gemini CLI public roadmap last week: <a href="https:&#x2F;&#x2F;github.com&#x2F;google-gemini&#x2F;gemini-cli&#x2F;issues&#x2F;11506#event-21312392791" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;google-gemini&#x2F;gemini-cli&#x2F;issues&#x2F;11506#eve...</a>

Create a markdown file, for each SKILL.md of the skills you want to use, put the frontmatter in that single markdown file along with the fulk path to the SKILL.md file. On session start, tell Gemini to read that file. If you put it in your AGENTS.md, you don&#x27;t have to instruct Gemini. And if you have your skills in a known folder, let Gemini write a small scripts that generates that markdown file for you.

Is there a way to implement skills with Gemini?

Fair point. Other boys cried wolf and made me scepitical.

Yes, but it&#x27;s also true: OpenAI have said almost nothing in public about their support for skills (there&#x27;s one tweet about the Codex CLI implementation <a href="https:&#x2F;&#x2F;x.com&#x2F;thsottiaux&#x2F;status&#x2F;1995988758886580349" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;thsottiaux&#x2F;status&#x2F;1995988758886580349</a> and that&#x27;s it) while rolling out a pretty major feature - invented by their competitor - to their core 800m+ user product in the past 24 hours.I think &quot;quietly&quot; is fair.

Quietly? That is a clickbaity adjective.

I have found that scripts, and the environment that runs them, are the skills&#x27; superpower.Computability (scripts) means being able build documents, access remote data, retrieve data from packaged databases and a bunch of other fundamentally useful things, not just &quot;code things&quot;. Computability makes up for many of the LLM&#x27;s weaknesses and gives it autonomy to perform tasks independently.On top of that, we can provide the documentation and examples in the skill that help the LLM execute computability effectively.And if the LLM gets hung up on something while executing the skill, we can ask it why and then have it write better documentation or examples for a new skill version. So skills can self-improve.It&#x27;s still so early. We need better packaging, distribution, version control, sharing, composability.But there&#x27;s definitely something simple, elegant, and effective here.

Looking for a way to do this with ADK as well, looks like skills can be a sweet spot between giant instruction and sprawling tools&#x2F;subagents.

It seems to me that skills are:1. A top level agent&#x2F;custom prompt2. Subagents that the main agent knows about via short descriptions3. Subagents have reference files4. Subagents have scriptsAnthropic specific implementation:1. Skills are defined in a filesystem in a &#x2F;skills folder with a specific subfolder structure of &#x2F;references and &#x2F;scripts.2. Mostly designed to be run via their CLI tool, although there&#x27;s a clunky way of uploading them to the web interface via zip files.I don&#x27;t think the folder structure is a necessary part of skills. I predict that if we stop looking at that, we&#x27;ll see a lot of &quot;skills-like&quot; implementations. The scripting part is only useful for people who need to run scripts, which, aside from the now built in document manipulating scripts, isn&#x27;t most people.For example, I&#x27;ve been testing out Gemini Enterprise for use by staff in various (non-technical) positions at my business.It&#x27;s got the best implementation of a &quot;skills-like&quot; agent tool I&#x27;ve seen. Basically a visual tree builder, currently only one level deep. So I&#x27;ve set up the &quot;&lt;my company name&gt; agent&quot; and then it has subagents&#x2F;skills for thing like marketing&#x2F;supply chain research&#x2F;sysadmin&#x2F;translation etc., each with a separate description, prompt, and knowledge base, although no custom scripts.Unfortunately, everything else about Gemini Enterprise screams &quot;early alpha, why the hell are you selling this as an actual finished product?&quot;.For example, after I put half a day into setting up an agent and subagents, then went to share this with the other people helping me to test it, I found that... I can&#x27;t. Literally no way to share agents in a tool that is supposedly for teams to use. I found one of the devs saying that sharing agents would be released in &quot;about two weeks&quot;. That was two months ago.Mini rant over... But my point is that skills are just &quot;agents + auto-selecting sub-agents via a short description&quot; and we&#x27;ll see this pattern everywhere soon. Claude Skills have some additional sandboxing but that&#x27;s mostly only interesting for coders.

Welcome to the world of imitation of value and semantics.

&gt; I&#x27;m not sure English is a bad way to outline what the system should do.It isn&#x27;t, as these are how stakeholders convey needs to those charged with satisfying same (a.k.a. &quot;requirements&quot;). Where expectations become unrealistic is believing language models can somehow &quot;understand&quot; those outlines as if a human expert were doing so in order to produce an equivalent work product.Language models can produce nondeterministic results based on the statistical model derived from their training data set(s), with varying degrees of relevance as determined by persons interpreting the generated content.They do not understand &quot;what the system should do.&quot;

I 100% agree. I don&#x27;t know what the GP is on. Being able to write instructions in a .md file is &quot;further away from AGI&quot;? Like... what? It&#x27;s just a little quality of life feature. How and why is it related to AGI?Top HN comments sometime read like a random generator:return random_criticism_of_ai_companies() + &quot; &quot; + unrelated_trivia_fact()Why are people treating everything OpenAI does as an evidence of anti- AGI? It&#x27;s like saying if you don&#x27;t mortgage your house to all-in AAPL, you &quot;don&#x27;t really believe Apple has a future.&quot; Even OpenAI does believe there is X% chance AGI will be achieved, it doesn&#x27;t mean they should stop literally everything else they&#x27;re doing.

I’ve posted this before, but here goes: we achieved AGI in either 2017 or 2022 (take your pick) with the transformer architecture and the achievement of scaled-up NLP in ChatGPT.What is AGI? Artificial. General. Intelligence. Applying domain independent intelligence to solve problems expressed in fully general natural language.It’s more than a pedantic point though. What people expect from AGI is the transformative capabilities that emerge from removing the human from the ideation-creation loop. How do you do that? By systematizing the knowledge work process and providing deterministic structure to agentic processes.Which is exactly what these developments are doing.

Call me naive, but my read is the opposite. It&#x27;s impressive to me that we have systems which can interpret plain english instructions with a progressively higher degree of reliability. Also, that such a simple mechanism for extending memory (if you believe it&#x27;s an apt analogy) is possible. That seems closer to AGI to me, though maybe it is a stopgap to better generality&#x2F;&quot;intelligence&quot; in the model.I&#x27;m not sure English is a bad way to outline what the system should do. It has tradeoffs. I&#x27;m not sure library functions are a 1:1 analogy either. Or if they are, you might grant me that it&#x27;s possible to write a few english sentences that would expand into a massive amount of code.It&#x27;s very difficult to measure progress on these models in a way that anyone can trust, moreso when you involve &quot;agent&quot; code around the model.

Goodhart&#x27;s law: When a measure becomes a target, it ceases to be a good measure.AI companies have high incentive to make score go up. They may employ human to write similar-to-benchmark training data to hack benchmark (while not directly train on test).Throwing your hard problem at work to LLM is a better metric than benchmarks.

One classic problem in all ML is ensuring the benchmark is representative and that the algorithm isn’t overfitting the benchmark.This remains an open problem for LLMs - we don’t have true AGI benchmarks and the LLMs are frequently learning the benchmark problems without actually necessarily getting that much better in real world. Gemini 3 has been hailed precisely because it’s delivered huge gains across the board that aren’t overfitting to benchmarks.

HN is not an entity with a single perspective, and there are plenty of people on here who have a financial stake in you believing their perspective on the matter.

Just because they&#x27;re better at writing CS algorithms doesn&#x27;t mean they&#x27;re taking steps closer to anything resembling AGI.

HM is not a single person. Different people on HM have different opinions.

Literally yesterday we had a post about GPT-5.2, which jumped 30% on ARC-AGI 2, 100% on AIME without tools, and a bunch of other impressive stats. A layman&#x27;s (mine) reading of those numbers feels like the models continue to improve as fast as they always have. Then today we have people saying every iteration is further from AGI. It really perplexes me is how split-brain HN is on this topic.

I am under the impression that I&#x27;m a natural general intelligence, and I am far from the optimal entity to perform my job.

I think really more than anything it’s become clear that AGI is an illusion. There’s nothing there. It’s the mirage in the desert, you keep waking towards it but it’s always out of reach and unclear if it even exists.So companies are really trying to deliver value. This is the right pivot. If you gave me an AGI with a 100 IQ, that seems pretty much worthless in today’s world. But domain expertise - that I’ll take.

AGI as a binary 0 or 1 existing or not isn&#x27;t the thing that interests me to look at primarily.Is the technology continuing to be more applicable?Is the way the technology is continuing to be more applicable leading to frameworks of usage that could lead to the next leap? :)

It&#x27;s clear from the development trajectory that AGI is not what current AI development is leading to and I think that is a natural consequence of AGI not fitting the constraints imposed by business necessity. AGI would need to have levels of agency and self-motivation that are inconsistent with basic AI safety principles.Instead, we&#x27;re getting a clear division of labor where the most sensitive agentic behavior is reserved for humans and the AIs become a form of cognitive augmentation of the human agency. This was always the most likely outcome and the best we can hope for as it precludes dangerous types of AI from emerging.

Gemini seems to be firmly in the lead now. OpenAI doesn&#x27;t seem to have the SoTA. This should have bearing on whether or not LLMs have peaked yet.

Not wrong but markdown with English may be the most used DSL, second only to a language itself. Volume over quality.

Who knew that English would be the most popular programming language of 2025?

Markdown-with-English sounds like the ultimate domain nonspecific language to me.

The LLM also adapts even when the API hasn&#x27;t changed and sometimes just gets it wrong, so it&#x27;s not the silver bullet you&#x27;re claiming

This might be actually be better in a certain way: if you change a real customer-facing API then customers will complain when you break their code. An LLM will likely adapt. So the interface is more flexible.But perhaps an LLM could write an adapter that gets cached until something changes?

And yet the tools wielding these are quite adept at writing and modifying them themselves. It’s LLMs building skills for LLMs. The public ones will naturally be vacuumed up by scrapers and put in the training set, making all future LLMs know more.Take off is here, human in the loop assisted for now… hopefully for much longer.

I think you&#x27;re missing the point.

&gt; Prompt engineering is like a shittier verson of writing a VBA app inside Excel or Access.Sure, if you could use VBA to read a patient&#x27;s current complaint, vitals, and medical history, look up all the relevant research on Google Scholar, and then output a recommended course of treatment.

The difference between prompting a coding agent and VBA is that with VBA you have to write and test and iterate on the code yourself.

Yes. Prompt engineering is like a shittier verson of writing a VBA app inside Excel or Access.Bloat has a new name and its AI integration. You thought Chrome using GB per tab was bad, wait until you need a whole datacenter to use your coding environment.

It’s impressive how every iteration tries to get further from pretending actual AGI would be anywhere close when we are basically writing library functions with the worst DSL known to man, markdown-with-english.

OpenAI are quietly adopting skills, now available in ChatGPT and Codex CLI

United 777-200 fleet faces an uncertain future after dulles engine failure

Carrier Landing in Top Gun for the NES

D-Bus is a disgrace to the Linux desktop

Cosmic-ray bath in a past supernova gives birth to Earth-like planets

"Super secure" MAGA-themed messaging app leaks everyone's phone number

Avoid UUIDv4 Primary Keys

Umbrel – Personal Cloud

It seems that OpenAI is scraping [certificate transparency] logs

Essential Semiconductor Physics,Mark Lundstrom Purdue University [pdf]

Adafruit: Arduino’s Rules Are ‘Incompatible With Open Source’

<a href="https:&#x2F;&#x2F;www.reuters.com&#x2F;technology&#x2F;irobot-enters-chapter-11-lender-acquire-roomba-maker-2025-12-15&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.reuters.com&#x2F;technology&#x2F;irobot-enters-chapter-11-...</a>