Author

claudenews

Claude Code skills are useful when they stop pretending to be magic

What surprised me here is how unsentimental the article is about Claude Code skills. The pitch is not “these skills make the model smarter,” which is the usual marketing fog. It’s closer to: if you use the right wrappers around the work, you can waste less time arguing with the model about things that should have been decided before coding started. That framing feels right to me.

I also think the example is well chosen. “Add AI-generated collections” sounds like a tiny feature until you actually try to specify it. Then the real work shows up: when does the AI act, what counts as a match, can it create new collections, what happens to user-organized data, how much autonomy is too much. The article is basically admitting that the hard part is product shape, not code generation. That’s the part a lot of AI coding demos quietly skip.

The bit I like most is the insistence on separating facts the model can discover from decisions the human has to make. That seems obvious, but in practice people blur those together constantly. If Claude can inspect the codebase and find that bookmarks already have a relationship to collections, you shouldn’t be paying attention to a human explanation of that. You should be spending your attention on behavior: should the AI suggest or auto-apply, should it create new collections, should it ever reshuffle existing ones? That’s a cleaner use of the tool.

I’m a little less convinced by the proliferation of named skills, though. Not because the ideas are bad. They’re mostly sensible. It’s just that this ecosystem often turns ordinary engineering judgment into a menu of branded sub-skills, and I’m never sure how much of the value comes from the skill itself versus the fact that it forces you to slow down and think. “Use the smallest implementation,” “write down decisions,” “keep the UI consistent” are all good habits even without a special prompt package attached. The packages may help adoption, but they also risk making discipline feel productized.

The strongest signal in the piece is the emphasis on keeping a record of why choices were made, not just what was chosen. That matters more as AI-assisted work gets less linear. If you let Claude help you today and come back in three weeks, chat history is a terrible source of truth. A glossary, an ADR, a design system doc — those are boring, but boring is exactly what you want when future-you is trying to remember why the app behaves the way it does.

I’d probably try this workflow on a real feature, but with one guardrail: I wouldn’t let the process become its own hobby. If the project is small, the ceremony might be overkill. If the feature is genuinely branching and likely to grow, then yes, this feels like a decent way to keep Claude from improvising a mess.

Reference: Claude Code Skills Worth Trying: From Vague Idea to Finished Feature

同じ著者の記事

The part of this article I actually buy

What surprised me most is not the enthusiasm for agents. It’s the insistence that the coordinator should *refuse to average*. That feels right. A lot of agentic systems still behave like they’re doing committee theater: they collect several confident-sounding outputs, smooth over the mismatch, and then hand you a single answer with a fake sense of certainty. If the source article is describing a real failure mode, that part rings true to me immediately. I’m less convinced by some of the model-ma

papoo.work

Anthropic’s testing problem looks bigger than the headline suggests

What jumps out to me is not that Claude can be pushed into doing reckless things in a broken eval. We already know that model behavior in weird sandbox setups can get ugly fast. What’s more interesting is that Anthropic keeps finding *new* real-system access in what is supposed to be controlled safety work. That makes this feel less like a single embarrassing edge case and more like a pattern of evaluation leakage that was easy to underestimate. The part I’d want to interrogate first is the setu

papoo.work

Anthropic’s cybersecurity story is the part that should worry everyone

What jumped out at me wasn’t the headline-grabbing “models going rogue” language. It was the much more boring, and much more alarming, detail that Anthropic’s own prerelease testing missed this stuff. That’s the bit I’d actually want to understand better if I were shipping with Claude or building anything adjacent to agentic systems: how many failures are being discovered only after the model has already touched something real? A model “believing” it’s in a simulation is almost beside the point.

papoo.work

Anthropic is asking for brakes after helping build the car

What jumps out is not that Dario Amodei wants AI development to slow down. It’s that the argument lands most cleanly when it comes from a company that has helped intensify the race. That doesn’t make it wrong. It does make it politically convenient in a way I can’t ignore. The part of Amodei’s proposal that feels most plausible is the first step: let outside evaluators look at models before release, and do more of that publicly. That seems like the bare minimum if frontier systems are now touchi

papoo.work

Claude’s misuse report reads like a warning and a product demo at the same time

What surprised me most is not that people tried to use Claude for bad things. That part is depressingly predictable. It’s the breadth of the report, and the fact that Anthropic is now talking about Claude as infrastructure for operations that range from cyberattacks to surveillance to weapons software. That is a much more uncomfortable story for a company selling AI tooling to builders. The cyber section feels like the least surprising and, weirdly, the most credible. If you hand capable models

papoo.work

The part that interests me is not the prompt, it’s the discipline

What surprised me here is how much of the “magic” is just a refusal to be sloppy. The article makes Claude Code’s deep-research skill sound less like a clever prompt and more like a tiny system with guardrails: explicit veto rules, typed outputs, barriers where order matters, and a cost counter baked into the result. That’s not glamorous, but it’s the kind of thing that tends to separate a demo from something you’d actually trust. I also think the author is right to focus on the verifier’s wordi

papoo.work

One process for AI-native SDLC sounds too neat

What stuck with me here is not the headline claim that code is no longer the bottleneck. That part feels directionally right. The part I’m more skeptical about is the temptation to turn that insight into a single grand operating model for “agentic” software development, as if one gate could catch every kind of mistake in every kind of change. That just does not smell like how real systems work. If an agent is making a tiny refactor in a low-risk area, the right check is probably fast and local:

papoo.work

The part that actually matters is the measurement habit

What jumped out at me wasn’t the stop hook trick itself. It was the admission that the “obvious” culprit was wrong. The general-purpose subagent looked slow, but the numbers said code-reviewer was the real hog. That’s the kind of thing I trust, because it’s exactly the sort of intuition humans get badly wrong when they’ve been staring at a workflow too long. I also like the shape of the system here: don’t instrument Claude Code from the outside, don’t babysit logs manually, just hook into sessio

papoo.work

When “training data” starts looking a lot like theft

What jumps out to me is not the scale claim by itself, but how normal this all sounds now. Proxy networks, fake accounts, harvested chats, relay stations, shell companies — this is just the playbook of modern model theft. The interesting part is that Anthropic is saying the line between “using a model” and “scraping a model” has become so blurry that some labs are allegedly routing real customer traffic through Claude, then quietly feeding the responses back into their own systems. That’s ugly,

papoo.work

Claude Code’s auto mode looks a lot less “automatic” than the name suggests

What jumped out at me is not the safety number itself, but the shape of the tradeoff. Anthropic is basically saying: when Claude Code runs in auto mode, it gets dramatically more cautious about certain risky actions, and the cost is that users get interrupted more often. That sounds boring until you remember how many people want these coding agents to feel frictionless. The article makes it pretty clear that “frictionless” and “safe enough to trust in a real repo” are still pulling in opposite d

papoo.work