polish

name	polish
description	Code quality assurance with automatic fix iteration. Execute lint/format/test and automatically fix errors until all pass (max 3 attempts). [What] Detect project configuration (mise.toml, package.json), run format/lint/test, and iteratively fix errors [When] Use when: users mention "polish", "品質保証", "lint fix", "format fix", "quality check", or need automated code quality improvement with repeated fixes [Keywords] polish, quality assurance, lint, format, test, automatic fix, code quality
argument-hint	[--with-comments]
disable-model-invocation	true
user-invocable	true
allowed-tools	Bash, Read, Grep, Edit

Polish - Code Quality Assurance

Overview

Automatically polish code by executing lint/format/test and fixing errors iteratively until all checks pass (maximum 3 attempts).

Core Capabilities

• Auto-detect project configuration (mise.toml, package.json)
• Execute format → lint → test workflow
• Iterative error fixing (max 3 attempts)
• Optional comment cleanup (with --with-comments flag)

Workflow

1. Detect Configuration

   • Check for mise.toml tasks (format, lint, lint-fix, test, ci)
   • Check for package.json scripts (format, lint, lint:fix, test)
   • Prefer explicit format / lint / test tasks when they exist
   • If a repo has no standalone lint or test task but has ci, use ci as the verification fallback instead of failing immediately

2. Execute Format

   • Run mise run format or npm run format
   • Report formatted files

3. Execute Lint & Fix

   • Run lint command
   • If errors found, run lint-fix command
   • If errors remain, attempt manual fixes
   • If no standalone lint command exists, use the lint-equivalent checks bundled in mise run ci
   • Repeat until clean or max attempts reached

4. Execute Tests

   • Run test command if available
   • Fix test failures if detected
   • If no standalone test command exists, use the test-equivalent verification bundled in mise run ci
   • Repeat until passing or max attempts reached

5. Report Results

   • Display summary of all steps
   • Show total attempts and execution time

Comment Cleanup (Optional)

When --with-comments flag is provided:

Remove

• Comments that merely repeat code content
• Self-evident comments (e.g., "constructor" above constructor)
• Obvious statements

Preserve

• WHY explanations
• Complex business logic explanations
• TODO, FIXME, HACK markers
• Non-obvious behavior warnings
• Important context information

Usage

# Basic execution
/polish

# With comment cleanup
/polish --with-comments

Configuration Detection

mise.toml

[tasks]
format = ["prettier --write ..."]
lint = ["markdownlint ...", "prettier --check ..."]
lint-fix = ["markdownlint --fix ...", "prettier --write ..."]
test = ["jest"]

package.json

{
  "scripts": {
    "format": "prettier --write .",
    "lint": "eslint .",
    "lint:fix": "eslint --fix .",
    "test": "jest"
  }
}

Supported Projects

• JavaScript/TypeScript (ESLint, Prettier, Jest)
• Python (Black, Flake8, pytest)
• Go (gofmt, golangci-lint, go test)
• Rust (rustfmt, clippy, cargo test)
• Markdown (markdownlint, prettier)

Notes

• Maximum 3 iteration attempts for the overall polish run, not 3 attempts per sub-step
• Each step's success/failure is clearly reported
• If test command is not found, that step is skipped
• If only ci exists, use format first and then ci as the verification pass
• Default to English output unless the user or repository instructions require another language