Ask HN: What former "must-have" AI agent skills have you deprecated?

5 points by lpjstack ↗ HN
As LLMs get stronger, I'm pruning my agent skill library. I've removed multi-step workflows (like superpowers) and web wrappers (like Agent Browse) that base models now handle natively, which speeds up context and improves performance.

Are you doing similar cleanup? What former go-to tools or skills have you recently ditched?

11 comments

[ 0.25 ms ] story [ 19.0 ms ] thread
Code review - it's now baked in to Claude Code and Codex CLI.
Code review is a good example. I used to keep fairly detailed review workflows, but the built-in behavior is now good enough for most cases.
In my professional life I've removed everything besides MCPs that connect to our knowledge base, gitlab, and jira.

For personal projects I never used them.

It was never that hard to type a couple extra sentences and it always felt like people were using an LLM to create most of these skills/plugins... and if the LLM can generate them already then what good is it to keep more garbage in context?