ClaudeCodeMod

All shelves / Mods / Review

prompt-polish

By VedantAndhale

prompt-polish is a Claude Code review mod by VedantAndhale.

Rewrites the draft in the prompt box when you press Improve, with one model call, and does not send the rewrite.

Draws

Install

The marketplace name is claude-pro-kit.

In Claude Code
/plugin marketplace add VedantAndhale/claude-pro-kit
/plugin install prompt-polish@claude-pro-kit
Open the repo

It runs inside the Claude Code process. The side panel is only what the author's README names. Read the repo before you install, or run claude plugin validate on the plugin folder.

Version
0.1.3
License
MIT
Repo stars
10
Last change
2026-10-08

What it can reach

  • Draws in the session

This is the one mod in the kit that makes a model call, and only when you press it. In the desktop app Improve sits in the footer under the prompt box, beside the model and Send, with Undo after a rewrite. It stays up while Claude works. In the terminal the same buttons show in a row above the prompt while the box holds a draft. ctrl+x tab focuses the row, then i improves, u undoes, and c cancels a call in progress. /polish with a prompt does the same. Improve sends the draft to Opus at low effort, or to Haiku or Sonnet if /polish model picks that, with a bundled copy of the prompt-master rules by nidhinjs (MIT), and puts the rewrite back in the box. It does not send the rewrite: you read it, edit it, or press Undo to get your original back. Drafts under 5 words are not sent to that call, and a toast says so. File paths, code, commands, names, URLs, and exact numbers are kept word for word. If you type over the draft while the model runs, your newer words win and the rewrite is dropped. A toast shows the call's exact tokens as the API reported them. One measured press on Opus at low effort was 13,790 input tokens and 73 output tokens, answered in 2.2 seconds. Most of that input is the prompt-master rules. These calls count toward your plan's usage limits like any other request on that model. Haiku is the cheaper choice if you press often.

One plugin from VedantAndhale/claude-pro-kit. The root README says prompt-polish is the only mod in that kit that makes a model call, and this one makes that call. The plugin folder has no README. These lines are the repository README. The call uses your own Claude login, so this page does not list network access. grammar-fix replaces the prompt Claude reads and shows each change under the message. This one is the Improve button, and the rewrite stays in the box.

Demo

The prompt-polish section on the author's GitHub page
Screenshot of the author's GitHub page.

What it hooks into

  • session.startWhen a session starts
  • command.runWhen you run its slash command/polish
  • prompt.editWhen you edit the prompt
  • prompt.submitWhen you send a prompt
  • session.endWhen a session ends
  • ui.renderDraws into the sessionSessionModeAbovePrompt

Read from the on() calls in its hook modules. Open its source under Files to check what each one does.

Slash commands 1

  • /polishPrompt polish: /polish <prompt> rewrites it into the prompt box (one call, exact tokens shown) · /polish model haiku|sonnet|opus

Files

The author's source, 3 files. They load when you open them.

Facts

Author
VedantAndhale
Category
Review
Marketplace
claude-pro-kit
Repo
VedantAndhale/claude-pro-kit
Folder
plugins/prompt-polish
Version
0.1.3
License
MIT
Repo stars
10
Last change
2026-10-08
Language
TypeScript
Files
8

Other mods in Review

sermo-translatejhao0413By default rewrites each prompt into English before sending it, and draws the reply's translation under the English text.fanyiJrCx7scCBy default translates a prompt that is not already English into English before Claude reads it, and draws a translation under each reply.serpiffserpiffFor serpiff, a local code-review app, draws how far this worktree's review has come and opens that review in the app.coa-deskdavebreamFor Claude of Alexandria, biblical study skills, draws a grounding line under each answer. Passage text comes from coa.davebream.com.receipthoobnnDraws a one-line receipt of the files, lines, and commands from the turn that just ended.figuresnatsukiumDraws diagrams and LaTeX math from Claude's replies, and images from tool results, as pictures in the transcript, in kitty or Ghostty.screen-guarddanyuchnDraws black bars over names, organisations, secrets, and contact details in the transcript while you share your screen.md-renderdeepskyblue86Redraws each finished assistant message as colored markdown: headings, bold, italic, code, lists, and quotes.readable-layoutGarvitNangruDraws your messages and Claude's answers in labeled boxes, short notes as quiet lines, and a rule after every turn.asd-ste100JAICHANGPARKAfter each answer, draws that answer as a Simplified Technical English page in a side pane. The same install includes the STE skill.glossary-browserchicioHolds a commandRefuses a Markdown edit that uses a glossary Avoid word, and draws that repo's terms in a band above the prompt.hide-diffsgixxy22Swaps a long diff for a one-line summary of the file and the lines changed, and folds each finished shell command to one line.
All 145 in Review