Licence
MIT
Version
0.2.0
Deps
4
Size
135 kB
Vulns
0
Weekly
0
lumiflare-claude-context-generator
Any repository → Claude Code context files in seconds.
Auto-generate REPOSITORY_OVERVIEW.md, feature_list.md, CLAUDE.md, and agent definitions for Claude Code.
Features
- Framework-agnostic — Detects PHP/Laravel, JS/TS/Next.js, Python/Django, Go/Gin, Ruby/Rails, Rust/Axum, and more
- Static analysis — Scans directory structure, routes, controllers, models, services, configs
- Agent team scaffolding — Generates 9 specialized agent definitions (architect, reviewer, implementer, etc.)
- Hybrid approach — CLI generates 80% from static analysis, then
context-bootstrapagent refines with Claude's intelligence
Quick Start
# Run directly with npx
npx lumiflare-claude-context-generator init
# Or install globally
npm install -g lumiflare-claude-context-generator
ccg init
What It Generates
project/
CLAUDE.md ← Claude Code instructions (adapted to your stack)
docs/
REPOSITORY_OVERVIEW.md ← Architecture, tech stack, directory structure
DOCUMENT_GUIDE.md ← Documentation standards
feature/
feature_list.md ← Feature map from controllers/routes/models
.claude/
agents/
orchestrator.md ← Team workflow coordinator
architect.md ← Design agent
reviewer.md ← Code review agent
implementer.md ← Implementation agent
test-writer.md ← Test generation agent
bug-investigator.md ← Bug investigation agent
refactoring-specialist.md ← Refactoring agent
doc-generator.md ← Documentation agent
context-bootstrap.md ← Deep analysis agent (refines CLI output)
Usage
Basic
# Analyze current directory (default: Japanese)
ccg init
# Generate in English
ccg init --lang en
# Analyze a specific directory
ccg init /path/to/project
# Preview without writing files
ccg init --dry-run
# Overwrite existing files
ccg init --force
Options
| Option | Description |
|---|---|
-l, --lang <language> |
Language for generated context: ja (default), en |
-f, --force |
Overwrite existing files |
--skip-agents |
Skip generating .claude/agents/ |
--skip-docs |
Skip generating docs/ |
--dry-run |
Show analysis without writing files |
Two-Phase Workflow
Phase 1: CLI Static Analysis (automatic)
The CLI detects your tech stack and scans the codebase:
✔ Detected: TypeScript + Next.js
✔ Scanned: 24 controllers, 18 models, 67 routes, 12 services
✔ Generated 13 files
Phase 2: Claude Deep Analysis (manual)
Open Claude Code in your project and say:
コンテキストを詳細化して
The context-bootstrap agent will:
- Read your actual source code deeply
- Refine
REPOSITORY_OVERVIEW.mdwith architecture insights - Enhance
feature_list.mdwith business logic details - Adapt
CLAUDE.mdrules to your project's conventions - Customize agent definitions for your stack
Detection Support
| Language | Frameworks | ORM | Test |
|---|---|---|---|
| PHP | Laravel, Symfony, CakePHP, Slim | Eloquent, Doctrine | PHPUnit, Pest |
| JavaScript/TypeScript | Next.js, Nuxt, Express, Fastify, NestJS, Hono | Prisma, TypeORM, Sequelize, Drizzle, Mongoose | Jest, Vitest, Mocha, Playwright, Cypress |
| Python | Django, Flask, FastAPI | Django ORM, SQLAlchemy | pytest |
| Ruby | Rails, Sinatra | ActiveRecord | RSpec, Minitest |
| Go | Gin, Echo, Fiber | GORM | (built-in) |
| Rust | Actix Web, Rocket, Axum | Diesel, SeaORM | (built-in) |
| Java/Kotlin | (detected from manifest) | — | — |
License
MIT