phileeppov

CHEFKISS

A 💩 button for AI slop. Claude takes the hint.

WHAT IT DOES

Every answer in Claude Code gets three buttons: 🤌 for how it’s said, 🎯 for what it says, 💩 for AI slop. Select a sentence first to react to that part only (in the desktop app). Claude replies to each click with a one-line takeaway, and new sessions start with your latest ones, 15 at most, in every project.

Animated buttons in the desktop app, plain emoji in the terminal.

After a few reactions, the /chefkiss pane can sum you up: one line about you, up to five on how you like your answers. Cross out what it got wrong and post the rest. Forget all erases lessons, the summary, and reactions.

Why not a line in CLAUDE.md? You click in the moment instead of editing a file, you can react to a single sentence, and Claude’s reply shows right away whether it understood.

Lessons are stored on your machine, and the mod doesn’t save Claude’s answers. There’s no chefkiss server.

Each click costs one Sonnet call on your Claude Code account, about 500 tokens. Sum up is one more call, only when you press it. You can turn those calls off in Settings; details in what gets sent where.

INSTALL

At the prompt of a Claude Code terminal session:

/plugin install chefkiss --marketplace phileeppov/chefkiss

In the desktop app, where /plugin does not run: click + next to the prompt box, then Plugins, Add marketplace with phileeppov/chefkiss, then Add plugin.

Source on GitHub →

Made by Nikita Filippov ☺

Product designer.
Send your product or idea, get a video review →

Not affiliated with Anthropic.