PlugWiki

genshijin-review Skill for Claude: One-Line Code Review Comments

Makes Claude write PR review comments as one line each: the line number, the problem and the fix.

Installs
12K
on skills.sh
Stars
331
on GitHub
Updated
Oct 2026
last repo commit

genshijin-review makes Claude write each review comment in one line as L42: <problem>. <fix>. For several files it uses <file>:L<line>: .... A severity prefix marks each point: 🔴 バグ (bug), 🟡 リスク (risk), 🔵 nit or ❓ 質問 (question).

It removes hedging such as 'it seems' and 'perhaps', keeps exact line numbers and backticked symbol names, and asks for a concrete fix instead of 'consider refactoring'. It switches back to normal paragraphs for security findings, architecture objections and onboarding context. SKILL.md says it only reviews: it does not edit code, approve or request changes, or run linters.

Good for

  • Reviewers who want short, actionable PR comments
  • Japanese-speaking teams reviewing code in Japanese
  • Developers tired of hedged review comments

Install genshijin-review

Claude Code

  1. 1

    Copy this prompt.

    Install the genshijin-review skill in Claude Code. Run this command:
    
    /plugin install genshijin
    
    Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

The plugin installs all 7 skills in the repo, 3 subagents and two hooks (SessionStart and UserPromptSubmit) that make Claude reply tersely by default. Its manifest registers no MCP server. The README says the Claude Code plugin is published in the official plugin directory. The last method installs only genshijin-review, with none of the hooks.

From the project README. Source.

What's in the skill

When it loads

超圧縮PRレビューコメント。1行1指摘: 位置・問題・修正。前置き削除、シグナル優先。 日本語対応。「PRレビューして」「コードレビュー」「/review」「/genshijin-review」で起動。 プルリクエストレビュー時に自動起動候補。

Sections in SKILL.md

  1. ルール
  2. 例
  3. 自動明瞭化
  4. 境界

Key facts

Installs on skills.sh
11,951 (Nov 2, 2026)

What it includes

  • L<line>: <problem>. <fix>. with a file prefix for several files.

    One finding per line, no preamble.

    Try: “PRレビューして”

    One-line format

  • 🔴 bug, 🟡 risk, 🔵 nit and ❓ question.

    Used when severities are mixed. Questions are for real doubts, not suggestions.

    Severity prefixes

  • Drops hedging and line explanations; keeps line numbers, symbols and concrete fixes.

    A single short praise line at the top is allowed.

    What it removes and keeps

  • Uses normal paragraphs for security, architecture and onboarding comments.

    It returns to the short style right after.

    Auto-clarity cases

  • Reviews only; it does not edit code, approve, request changes or run linters.

    Output is formatted to paste into a PR.

    Boundary

From the official docs

Permissions

Low risk

The skill only writes review text and bundles no scripts. The plugin route adds hooks that change how Claude replies in every session, so use the skills CLI command if you want only this skill.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • The plugin changes every session

    The plugin route registers a SessionStart and a UserPromptSubmit hook that run node scripts. They switch Claude into terse genshijin replies by default, write a flag file under ~/.claude and, if you have no statusLine, tell Claude to suggest adding one to settings.json. Set GENSHIJIN_DEFAULT_MODE=off to stop the auto start. github.com

Questions

What does genshijin-review do?

It makes Claude write PR review comments one line at a time, each with the line number, the problem and the fix. Severity is marked with a bug, risk, nit or question label.

How do I install genshijin-review?

Use the install box above. The README route is the genshijin plugin, which adds 6 other skills, 3 subagents and hooks. The skills CLI command with --skill installs only this skill.

Does it approve or fix the code?

No. SKILL.md says it only reviews. It does not edit code, approve, request changes or run linters.

Does it work in English?

SKILL.md is written in Japanese and its description says it supports Japanese. It does not state an English output mode.

When does it write longer comments?

For security findings, architecture objections that need reasoning, and comments for new team members who need the why. It returns to one-line comments afterwards.

Is it safe to install?

The skill folder is one markdown file that only produces review text. The plugin route registers hooks that run node scripts at session start and on every prompt. Use the single-skill command to avoid them.

Information

updated Nov 2, 2026

Folder
skills/genshijin-review

interfacex-co-jp/genshijinskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 6
  1. genshijin README · github.com
  2. plugin.json · github.com
  3. genshijin-activate.js · github.com
  4. genshijin-review SKILL.md · github.com
  5. genshijin-review on skills.sh · skills.sh
  6. InterfaceX-co-jp/genshijin · github.com

Updated Nov 2, 2026.