buckle-up

KO | EN

gitlite — search

#javascript #ai #typescript #llm #golang #machine-learning #rust #deep-learning #docker #java #security #ios #openai #open-source

(★ 10)

Claude Code plugin: Project setup wizard using your research catalogue

buckle-up 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# buckle-up

> Strap in — we're configuring this project properly.

**buckle-up** is a Claude Code plugin that uses your research catalogue to configure the optimal toolset for any project.

## The Problem

You've researched 30+ AI coding tools. Now you're starting a new project and need to pick the right combination. Do you:

A) Spend an hour re-reading all your notes
B) Install everything and mass-produce ~/.claude.json backups
C) Run `/buckle-up` and let it figure it out

## How It Works

```mermaid
flowchart TD
 A["/buckle-up"] --> B{Path provided?}
 B -->|yes| C{Index stale?}
 B -->|no| B1{Choose catalogue}
 B1 -->|"Use existing"| B2[Enter path]
 B1 -->|"Clone readymade"| B3["git clone razpetel/research-catalogue"]
 B2 --> C
 B3 --> C
 C -->|"hash changed"| D["Rebuild index (shows progress)"]
 C -->|"hash matches"| E{Previous run?}
 D --> E
 E -->|yes| F{Re-run mode?}
 E -->|no| G[Detect existing config]
 F -->|Resume| L1
 F -->|Upgrade| G
 F -->|Re-interview| G
 F -->|Reset| Z[Remove config]
 G --> H["Interview (5-10 questions)"]
 H --> I[Score all tools]
 I --> J["Top 10 candidates (category coverage)"]
 J --> K["Present plan [Apply|Edit|Explain]"]
 K -->|approve| R["🔍 research-agent (fetch latest details)"]
 R --> L1["Apply MCPs + Plugins"]
 L1 --> L2["📝 config-writer-agent (generate CLAUDE.md + hooks)"]
 L2 --> V["✓ verification-agent (validate config)"]
 V -->|pass| M[Done]
 V -->|issues| FIX[Fix issues]
 FIX --> V
 Z --> M
```

**The key steps:**

1. **Index** — Reads your research catalogue, extracts structured data (cached)
2. **Interview** — Asks 5-10 smart questions about your project
3. **Score** — Ranks every tool against your specific needs
4. **Research** — Agent fetches latest details for approved tools
5. **Configure** — Agent generates CLAUDE.md + hooks using best practices
6. **Verify** — Agent validates everything works before declaring done

## Prerequisites

A research catalogue with:
- `catalogue.md` — Index of researched tools
- `catalogue/*.md` — Individual research reports

Don't have one? Two options:
- **Quick start:** Clone the [readymade catalogue](https://github.com/razpetel/research-catalogue) (buckle-up offers this automatically)
- **Build your own:** Install [fomo-researcher](https://github.com/razpetel/fomo-researcher) and run `/research` on tools you care about

## Insights Integration

Buckle-up automatically gathers usage patterns from `/insights` to personalize interview questions. If no recent insights exist, it offers to generate them.

## Installation

```bash
/plugin marketplace add razpetel/buckle-up
/plugin install buckle-up@razpetel
```

Restart Claude Code. (Yes, really.)

## Usage

```bash
# Choose catalogue interactively (use existing or clone readymade)
/buckle-up

# Explicit catalogue path
/buckle-up /path/to/catalogue.md
```

## What Gets Configured

| Component | How | Scope |
|-----------|-----|-------|
| **MCPs** | Added to `~/.claude.json` | Global |
| **Plugins** | Installation commands provided | Global |
| **CLAUDE.md** | Methodology references added | Project |
| **Hooks** | Test gates, TDD reminders | Project |

**Note:** MCPs and plugins are global — they persist across projects. Running buckle-up on multiple projects accumulates them. This is intentional (most tools are useful everywhere).

## Re-runs

Changed your mind? Discovered a shiny new tool? Run `/buckle-up` again to:

- **Resume** — Finish interrupted setup
- **Upgrade** — Check for new tools in catalogue
- **Re-interview** — Start fresh with different needs
- **Reset** — Remove project config (global MCPs/plugins remain)

## State

- Configuration saved to `.claude/buckle-up-state.json`
- Backups at `.claude/buckle-up-backup/`
- Index cached at `{catalogue-dir}/.toolshed-index.json`

## How Scoring Works

Each tool is scored across 5 dimensions:

| Dimension | What It Measures |
|-----------|------------------|
| Team fit | Solo vs team, matches your situation |
| Maturity | Production-ready vs beta |
| Complexity | Minimal vs feature-rich, matches your preference |
| Token efficiency | Context usage |
| Autonomy | Human-in-loop vs overnight autonomous |

Weights adjust based on your answers. Bonuses for complementary tools, penalties for conflicts.

## Agents

Buckle-up delegates specialized tasks to agents:

| Agent | When | What It Does |
|-------|------|--------------|
| **research-agent** | After you approve | Fetches latest versions, install commands, breaking changes |
| **config-writer-agent** | During apply | Generates CLAUDE.md and hooks using best practices |
| **verification-agent** | After config | Validates everything works before declaring done |

The agents respect your preferences:
- Token-conscious? Minimal CLAUDE.md (<500 words)
- High autonomy? Proactive hooks, less confirmation
- Production project? Test gates and quality checks

## License

MIT

---

*Built because manually configuring Claude Code for every project is not a personality trait.*

buckle-up

buckle-up 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

buckle-up

buckle-up 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README