We ship more code with AI than ever. And explain less of it.
So I built grip: a git hook that quizzes you on your own diff before you push. 5 questions, score out of 100, under the pass mark β blocked.
Pop quiz from your own repo. π
One curl to install. Claude Code plugin too.
https://t.co/SGkYfJn1h9
#agentic #code #claude #cursor #ai
Try it in any agent:
npx skills add guilyx/grip -g
Open source, BSD-3:
https://t.co/YYhBEAWMG1
Docs and what we haven't measured yet:
https://t.co/GqEbMSC31l
grip now lives inside your coding agent.
Type /grip before you push: 5 questions on the diff the agent just wrote. You answer, not the agent. Below the pass mark, no push.
We let AI write the code. We still own it.
Links below π
Grip aims at reducing the amount of "let me check" replies you get from the devs in your team by installing a git hook that quizzes you on your own diff before it goes upstream. Before anyone else has to.
I only keep a grip on my 9-5 code, anything else is out of hand.
βοΈ Runs on Claude Code, Codex or Gemini CLI (your existing sub), an API key, or local Ollama
β‘ One curl to install. No Python, no pip
π Also a Claude Code plugin: /grip:quiz inside the agent, plus a hook so Claude can't push a diff you haven't passed
π https://t.co/XzWbPzURIf
π https://t.co/obCfpFvf2E
Keep a grip on your code! π€
would be curious to run a grip webapp where people get paid for a limited amount of time to hook it up and give us the result...
goal would be to understand with 100, 1000 people how much they score on average
but how do I prevent them from using an LLM to pretend they have grip ?
https://t.co/XzWbPzURIf
would be curious to run a grip webapp where people get paid for a limited amount of time to hook it up and give us the result...
goal would be to understand with 100, 1000 people how much they score on average
but how do I prevent them from using an LLM to pretend they have grip ?
https://t.co/XzWbPzURIf
@typesafeai trying to build some applications for robotics, research.. is there any way to fast track getting an account through a research institute ?
Received good feedback on grip. People from my team tried it out, spreading through the company now.
Looking for suggestions!
π Code: https://t.co/SGkYfJnz6H
π Docs: https://t.co/WMLYXaIjBk
π Claude Code plugin: /plugin marketplace add guilyx/grip
β‘ Install: curl -fsSL https://t.co/KN2rrrVqVi | sh
BSD licensed. Issues and PRs welcome.
"Why does the retry return early before the balance check?"
Fair question in review. The author pushed it two days ago and now has to reread their own diff to answer. We've all been that author. π
The agent knew. We didn't. And we're the ones pushing to main.
grip moves that check to before the PR exists. You answer for your diff before anyone else has to. Reviewers get a change the author can actually explain. Everyone saves time.
AI makes code cheap to write. Owning it is still on us. π§
#ai #agentic #claude #cursor
We ship more code with AI than ever. And explain less of it.
So I built grip: a git hook that quizzes you on your own diff before you push. 5 questions, score out of 100, under the pass mark β blocked.
Pop quiz from your own repo. π
One curl to install. Claude Code plugin too.
https://t.co/SGkYfJn1h9
#agentic #code #claude #cursor #ai