kage-bunshin
CLAUDE.md1 starsChanged 8 months agoMITView on GitHub
An OSS agent system for Claude Code where multiple specialized agents collaborate on intellectual work beyond coding.
In this repository5 other files it gives its agents.
- brainstormSkill
- evaluateSkill
- learningSkill
- notebookSkill
- plan-projectSkill
The file
We split this CLAUDE.md into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 77
When the user's request matches one of the following patterns, automatically invoke the corresponding skill — even without an explicit …
- Distinctive: line 87
If Notebook is not configured, skip persistence and output results only
- Distinctive: line 91
All Notebook writes must include structured frontmatter
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Kage Bunshin - Second Brain Agent System
An OSS agent system for Claude Code where multiple specialized agents collaborate on intellectual work beyond coding.
Architecture
Agents (.claude/agents/)
4 specialized agents. Each has a unique thinking style and writes its own deliverables.
- ideator - Divergent thinking, brainstorming, creative idea generation
- analyst - Critical analysis, risk assessment, quality reviews, feasibility evaluation
- planner - Task breakdown, scheduling, dependency management
- researcher - Deep investigation, information gathering, multi-source verification
Skills (.claude/skills/)
Workflows that create Agent Teams for iterative multi-agent collaboration.
/brainstorm- Iterative ideation via ideator ⇄ analyst team/evaluate- Investigation and assessment via researcher ⇄ analyst team/plan-project- Research-backed planning via researcher ⇄ planner team/learning- Evidence-based retrospective via researcher ⇄ analyst team/notebook- Notebook search and management
Notebook
A separate repository for storing user-specific data. All deliverables are structured as Markdown with frontmatter.
Notebook Setup
// .claude/settings.local.json
{
"env": {
"NOTEBOOK_PATH": "/path/to/your/notebook-repo"
},
"permissions": {
"additionalDirectories": ["/path/to/your/notebook-repo"]
}
}NOTEBOOK_PATH: Absolute path to the Notebook repositoryadditionalDirectories: Required because the Notebook is an external repository outside Claude Code's default file access scope
Notebook Directory Structure
notebook-repo/
├── ideas/ # ideator
├── reviews/ # analyst
├── research/ # researcher
├── plans/ # planner
└── learnings/ # analyst via /learningNotebook Entry Format
---
title: Entry name
type: idea | review | research | plan | learning
date: YYYY-MM-DD
project: Project name
tags: [tag1, tag2]
related: [path/to/related.md]
---
BodyLanguage Configuration
// .claude/settings.local.json
{
"env": {
"OUTPUT_LANGUAGE": "en"
}
}OUTPUT_LANGUAGE: Set the output language (en,ja, etc.). If not set, agents match the user's input language (default: English).
Multi-Agent Routing
When the user's request matches one of the following patterns, automatically invoke the corresponding skill — even without an explicit slash command:
- Brainstorming, idea generation, "any good ideas?" →
/brainstorm(ideator + analyst) - Feasibility assessment, "is this viable?", "should we do this?" →
/evaluate(researcher + analyst) - Project planning, task breakdown, "make a plan" →
/plan-project(researcher + planner) - Retrospective, reflection, "what did we learn?" →
/learning(researcher + analyst) - Notebook search, "find past ideas", "show saved content" →
/notebook
Rules
- At the start of a new session, if
NOTEBOOK_PATHis set, scan the Notebook for the 3 most recently updated entries (across all directories, based on modification time or thedatefield in frontmatter) to understand the user's current context and ongoing work - Each agent writes its own deliverables
- If Notebook is not configured, skip persistence and output results only
- Each agent does not work outside its area of expertise
- Skills always operate through Agent Teams with iterative multi-agent collaboration
- Output language follows
OUTPUT_LANGUAGEsetting, or matches the user's language (default: English) - All Notebook writes must include structured frontmatter
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

