Genshijin Crew (subagent delegation guide)

A decision guide for delegating code lookup, small edits, and diff reviews to compressed-output subagents so your Claude Code context lasts far longer.

Dev & CodingAdvanced★ 332⑂ 14AI score 9/10Last updated: Aug 15, 2026

What it does

Gives a decision table for when to use and when to avoid three subagents: genshijin-investigator (locate code), genshijin-builder (edit 1-2 files), genshijin-reviewer (review a diff).
Since subagent tool-results are injected verbatim into the main context, this preset compresses their output ("caveman style"), cutting roughly 60% of tokens — e.g. vanilla Explore returning 2k tokens vs. ~700 tokens from the investigator.
Locks in explicit output contracts: investigator returns path:line — \symbol` — note(greppable viapath:\d+), builder returns a change range plus a verified:line or a terminal token liketoo-big., reviewer returns path:line: severity: issue. fix.` plus totals.
Provides chaining patterns: locate → fix → verify, parallel scouting with 2-3 investigators in one message, and single-shot edits when the site is already known.
States hard prohibitions: never call builder without an identified file, never chain for a 5-file refactor, never ask the reviewer for "general feedback", and don't expect prose.

Who it's for

Developers on large codebases who keep hitting context exhaustion mid-session.
Refactoring-heavy workflows that repeat search → edit → review dozens of times per session.
Users already on subagents but unsure when vanilla Anthropic agents are the better call.

Examples

"Find every reference to AuthToken" → investigator returns a compact path:line list; the main thread picks 1-2 sites and hands paths to the builder.
"Audit the diff on this branch for bugs" → reviewer returns severity-tagged findings sorted by file then line, with a totals row.
Broad investigation → launch 3 investigators in parallel from one message (definitions vs. callers vs. tests) and aggregate in the main thread.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/InterfaceX-co-jp/genshijin/HEAD/skills/genshijin-crew/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)

Download the ZIP with the button below.
In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.

↓ Download ZIP

Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/genshijin-crew folder from the GitHub repo InterfaceX-co-jp/genshijin into my ~/.claude/skills/genshijin-crew/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/InterfaceX-co-jp/genshijin.git && mkdir -p ~/.claude/skills && cp -r genshijin/skills/genshijin-crew ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

Open a terminal.
Clone the repo: git clone https://github.com/InterfaceX-co-jp/genshijin.git
Create the skills folder: mkdir -p ~/.claude/skills
Copy the skill: cp -r genshijin/skills/genshijin-crew ~/.claude/skills/
This guide assumes the genshijin-investigator, genshijin-builder, and genshijin-reviewer subagents exist. Check the repo's agents/ directory (or same-named skills) and copy them into ~/.claude/agents/ or ~/.claude/skills/ as well.
Restart Claude Code and try a prompt like "delegate with genshijin-crew" or "investigate this in a context-saving way" to confirm it triggers.

View source on GitHub ↗License: MIT