AI 어시스턴트용 AGENTS.md 템플릿

이 템플릿을 사용하여 AI 코딩 어시스턴트(Claude Code, Cursor, Cline, Windsurf)를 자율적인 Rediacc 인프라 관리에 맞게 구성하세요. 아래 블록을 프로젝트의 CLAUDE.md, .cursorrules 또는 그에 상응하는 에이전트 설정 파일에 복사하세요.

템플릿

# Rediacc Infrastructure

Self-hosted infrastructure platform. Each machine runs Docker-based repositories with encrypted, isolated environments.

## CLI Tool: rdc

Install: `curl -fsSL https://www.rediacc.com/install.sh | bash`

### Key Flags
- `--output json` (or `-o json`), machine-readable JSON output
- `--yes` (or `-y`), skip interactive confirmations
- `--quiet` (or `-q`), suppress informational output
- `--fields name,status`, limit output fields
- Auto-JSON: when piped (non-TTY), output defaults to JSON automatically

### JSON Envelope
All JSON output uses a consistent envelope:
{"success": true, "command": "...", "data": ..., "errors": null, "warnings": [], "metrics": {"duration_ms": N}}

On error: {"success": false, "command": "...", "data": null, "errors": [{"code": "...", "message": "...", "retryable": false, "guidance": "..."}], ...}

### Common Operations

# Machine status
rdc machine query --name <machine> -o json

# List containers on a machine
rdc machine containers --name <machine> -o json

# Machine health check
rdc machine health --name <machine> -o json

# Deploy a repository
rdc repo up --name <repo> -m <machine> --yes

# Stop a repository
rdc repo down --name <repo> -m <machine> --yes

# SSH terminal to machine
rdc term connect -m <machine>

# SSH terminal to specific repo (sets DOCKER_HOST)
rdc term connect -m <machine> -r <repo>

# Run command on machine
rdc term connect -m <machine> -c "command"

# File sync
rdc repo sync upload -m <machine> -r <repo> --local ./local-path
rdc repo sync download -m <machine> -r <repo> --local ./local-path

# List all available commands with schemas
rdc agent capabilities

# Show schema for a specific command
rdc agent schema --command "machine query"

### Architecture
- **Repository**: Isolated application deployment with its own Docker daemon at /var/run/rediacc/docker-<networkId>.sock, loopback IP range (127.0.x.x/26), and encrypted btrfs mount at /mnt/rediacc/mounts/<guid>/
- **Config**: CLI config at ~/.config/rediacc/rediacc.json. Auto-created on first use.
- Two adapters: **local** (default, SSH-based) and **cloud** (experimental, API-based)

### Rules
- Use "local adapter" / "cloud adapter", never say "modes"
- S3 is a resource state backend, not a separate adapter
- Default config is created automatically on first use, do not tell users to run `rdc config init`
- Always use `--output json` when parsing output programmatically
- Always use `--yes` to skip confirmations in automated workflows
- Use `--dry-run` on destructive commands (repo delete, repo up, repo down) to preview before executing

커스터마이징

<machine>과 <repo>를 실제 머신 및 리포지토리 이름으로 교체하세요. rdc config repository list를 실행하면 이름과 GUID 매핑과 함께 사용 가능한 리포지토리 목록을 확인할 수 있습니다.

설정 확인

# 구성된 머신 목록 조회
rdc machine query --name <machine-name>

# GUID와 함께 리포지토리 목록 조회
rdc config repository list

# 사용 가능한 명령어 확인
rdc agent capabilities

에이전트별 설정

• Claude Code: 프로젝트 루트에 CLAUDE.md로 저장
• Cursor: 프로젝트 루트에 .cursorrules로 저장
• Cline: Cline 시스템 프롬프트 또는 프로젝트 지침에 추가
• Windsurf: 프로젝트 루트에 .windsurfrules로 저장