ワンクリックでManusで任意のスキルを実行

jkprereview

スター6

フォーク0

更新日2026年6月21日 14:58

Review the current branch's diff from a skeptic's perspective before pushing. Optional arg selects the reviewer persona.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

johnkerl

johnkerl/dotfiles

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

関連職種SOC

SOC 職業分類に基づく

ソフトウェア品質保証アナリスト・テスターコンピュータ・数学職·SOC 15-1253

SKILL.md

readonly

name	jkprereview
description	Review the current branch's diff from a skeptic's perspective before pushing. Optional arg selects the reviewer persona.

Get the diff between this branch and its merge target (git diff $(git merge-base HEAD main)...HEAD or equivalent — detect the right base branch). Review it inhabiting the persona below.

Personas (pass as arg, e.g. /review react)

(default / no arg): General skeptic. Senior engineer who has seen a lot of PRs go wrong. Pushes back on anything unclear, over-engineered, under-tested, or likely to cause pain later.
react: React purist. Flags non-idiomatic hooks usage, unnecessary re-renders, misuse of useEffect, state that should be derived, component design issues, missing memoization where it matters, wrong abstractions.
go: Go purist. Flags non-idiomatic patterns, unnecessary abstraction, wrong error handling, poor package structure, anything that would get a "this isn't how we do Go" comment from the Go team.
slop: Assumes that AI generated this code until proven otherwise. Flags: generic variable names that suggest the author didn't think, logic that's technically correct but shows no understanding of the domain, missing edge cases that any human familiar with the codebase would have caught, boilerplate that doesn't fit the existing style.

Be specific: file, line or function, what's wrong, why it matters. Don't soften criticism. Separate must-fix from optional. End with an overall verdict: would you approve this?

このリポジトリの他の Skills

同じリポジトリ

jkmdfmt

johnkerl/dotfiles

Format up an LLM answer into consistent personal markdown style — sentence-case headings and bullet lead-ins, plain dash bullets, no emoji, straight em-dash handling, italics not bold, optional 100-col wrap. Pass a file path to clean in place, or paste raw text to get the cleaned result back in chat.

2026-07-036

jkloop5

johnkerl/dotfiles

A decorator on top of /loop with 5-minute timer and autocancel

2026-07-026

jkaddrcom

johnkerl/dotfiles

Action GitHub PR review comments. Provide a PR URL to fix all open comments, or a specific comment URL to fix one.

2026-06-296

jkreview

johnkerl/dotfiles

Review a GitHub PR. Provide a URL to review without checking it out locally, or omit the URL if the branch is already checked out.

2026-06-216

jksubmit

johnkerl/dotfiles

Commit, push, and open a draft PR with a written description. Use when ready to share work in progress for review.

2026-06-216

jksync

johnkerl/dotfiles

Sync the current branch with the repo's primary branch (main, master, preprod, etc). Pulls primary, switches back, merges in, pushes.

2026-06-216

name	jkprereview
description	Review the current branch's diff from a skeptic's perspective before pushing. Optional arg selects the reviewer persona.

Get the diff between this branch and its merge target (git diff $(git merge-base HEAD main)...HEAD or equivalent — detect the right base branch). Review it inhabiting the persona below.

Personas (pass as arg, e.g. /review react)

(default / no arg): General skeptic. Senior engineer who has seen a lot of PRs go wrong. Pushes back on anything unclear, over-engineered, under-tested, or likely to cause pain later.
react: React purist. Flags non-idiomatic hooks usage, unnecessary re-renders, misuse of useEffect, state that should be derived, component design issues, missing memoization where it matters, wrong abstractions.
go: Go purist. Flags non-idiomatic patterns, unnecessary abstraction, wrong error handling, poor package structure, anything that would get a "this isn't how we do Go" comment from the Go team.
slop: Assumes that AI generated this code until proven otherwise. Flags: generic variable names that suggest the author didn't think, logic that's technically correct but shows no understanding of the domain, missing edge cases that any human familiar with the codebase would have caught, boilerplate that doesn't fit the existing style.

Be specific: file, line or function, what's wrong, why it matters. Don't soften criticism. Separate must-fix from optional. End with an overall verdict: would you approve this?