Top
Best
New

Posted by domhudson 20 hours ago

I-have-ADHD: A skill to stop coding agents from burying the answer(github.com)
451 points | 310 commentspage 2
menthe 2 hours ago|
Seems everyone on Hacker News has already forgotten about Anthropic’s amazing new watermarking algorithm. You don’t need to look any further to understand why it writes so awfully - it literally swaps words programmatically all over the text from the result of a hash function…
pona-a 2 hours ago|
This was an issue before watermarking was publically announced. I'm also not an expert, but aren't bits of a hash indistinguishible from uniform randomness?
LoganDark 2 minutes ago||
Depends on the particular hash. Some hashes are designed to produce similar outputs when the input data is similar (for example, SSDeep). I'd imagine hashes designed to survive light copywriting would also not be entirely indistinguishable from random.
andai 15 hours ago||
>explain like I have executive dysfunction

This is true. Transformer has several orders of magnitude more working memory than any human. Compared to transformer we all have executive dysfunction.

By default they explain things assuming I have infinite processing bandwidth. I do not! I have several zeroes less than they do.

beckhamc 14 hours ago|
Pretty much this. I actually find this more annoying than the way it writes, but that is also annoying as well.
codetiger 2 hours ago||
Recent language Claude uses is very heavily loaded and hurts me when reading. So I had to use this to keep my sanity. This started only with Opus 5 where Anthropic thought they optimized the meaning per token and make it very hard to read. Comments and code also looks like cryptic.
mcv 3 hours ago||
Sounds useful. I'm often struggling with how extremely verbose Claude is. Shorter is better. Cut the unnecessary prose and the Claudian turns of phrase.

But this seems to focus very much on telling the user what to do, whereas usually I'm telling Claude what to do. It feels like this inverts the relationship and wants to turn me into a reverse centaur.

Wouldn't "lead with the answer" be better than "lead with the action"? But sometimes answers do require detail and explanation. I just want to get rid of all the unnecessary prose.

al_borland 13 hours ago||
I’ve been scolding my LLM quite a bit lately for this. Instead of telling me what to do, it first tells me 3 ways that might technically work, but I probably shouldn’t use for various reasons. This is all a waste of my time, tokens, and sanity.
NewJazz 13 hours ago|
They bill per token, what do you expect?
gerad 19 hours ago||
Personally I ask it to explain it simply like it’s talking to an executive reading it on their phone while on another meeting.
mjsarfatti 15 hours ago||
You don’t need a skill. Just tell it to BLUF (bottom line up front) and ASD-STE100.

And stop using Opus 9 Pro Max XHigh 10.0 for everything. If you choose a hyper-thinking model for asking the weather you can’t but expect yapping.

0xffff2 15 hours ago||
How do you actually incorporate these into your prompt? Because I've tried and I've seen many others try and report that it does not work well, especially over long sessions.

I also don't know how much to trust the model, but I've had the model tell me specifically that certain aspects of ASD-STE100 are unactionable and will just create more noise.

The OP's own skill even leads with something in a very similar vein:

> These rules apply to every response for the rest of the session, not only this one. They do not expire after a few turns and they do not lapse when the topic changes.

My understanding is that phrases like this are at best a _very_ weak signal to the model. It's simply contradictory to how the model works at a level that can't be overridden by injecting tokens.

mjsarfatti 3 hours ago||
Weak is even giving it too much credit imho, a yapping skill telling Claude to not yap can’t but fail.

I placed those instructions in setting > personal instructions, in my global CLAUDE.md, and in each project’s CLAUDE.md. I also use the concise output style in CC. If I choose a high thinking model it will start deviating in long sessions, then I just remind it in my next message:

“Remember ASD-STE100 style.

[rest of my message]”

BLUF sticks a lot easier than STE to be honest… but Claude knows what STE is, and using the “ASD-STE100 style” locution avoids the compliance issue (it’s true that strict ASD-STE100 compliance isn’t really possible, nor desired)

code_biologist 12 hours ago||
There's no physical reason "smartest/most capable model" should be "most yappy model". Smart people can write concisely. It's unfortunate you have to tolerate/fight yappiness to get intelligence with current SotA models.
hbbio 6 hours ago||
Just in case the model has ADHD too, I am using a minified version of it:

https://gist.github.com/hbbio/2faf096cbb77e197233ab9a2958beb...

whirlwin 19 hours ago||
Can someone explain why this is a agent skill and not just agent config (AGENTS.md / CLAUDE.md)?
SamuelAdams 18 hours ago||
Skills are kind of a new thing starting around January 2026, but the spec was established about 12 months ago.

https://news.ycombinator.com/item?id=46871173

Anyways the most layman way I’ve seen it explained is this: skills help save token usage for the right context. Not every request needs all instructions all the time - running tests is different than reviewing a PR, so why should the context window have instructions for both on every request?

So now you split instructions into “skill” files, which are basically opinionated markdown files. And you invoke those with something like /grill-me in the prompt depending on what you’re doing.

There are some steps to have the agent automatically know what to invoke for you but in my experience this automation is hit or miss.

It is also challenging to keep track of a growing library of skills and keeping those up to date.

So YMMV regarding skills. I typically keep things in a single markdown file even if the context window gets a bit bloated.

whirlwin 15 hours ago||
Yeah, I get that. I have created dozens of skills, but they are usually to accomplish very specific tasks. Some of them have accompanying scripts, etc.

My point is that this sounds more like a general AGENTS.md use case, similar to defining tone of voice, output format, etc.

It just seems skills is the only way to distribute certain "behavior" as of today. But not everything is a skill IMHO, and not this is not one of them.

On the other hand, maybe we're seeing an evolution of what skills are becoming.

iambenm 18 hours ago|||
You can /i-have-adhd when you get a wall of text and are exasperated. You don't always want the responses that it generates when you /i-have-adhd, and things in AGENTS.md tend to lose weight as context grows.
mjsarfatti 3 hours ago|||
Because it’s wrong, and you are right. It should be in your global AND per-project md’s… Perhaps in a custom output style too.
trio8453 19 hours ago|||
It's a fashion choice.
drdexebtjl 18 hours ago|||
It lets the author pretend to be doing software engineering, like running tests and deploying versions.
throwaway613746 18 hours ago||
[dead]
abathologist 18 hours ago|
"Software Engineering" in 2026.
neutrinobro 15 hours ago||
Can't read reads walls of text in Claude-ish prose? Sounds to me like you just have a SKILLS.md issue.
rbtprograms 17 hours ago||
"AGI is here"
More comments...