hungrytech-claude-skills
A curated collection of 12 Claude Code plugins that form a complete software engineering team. Each plugin is a specialized expert agent — no runtime build system, just shell scripts and SKILL.md prompts.
한국어 버전은 아래에 있습니다 →
Team Overview
User Request
│
▼
sub-team-lead (classify & route)
├── Single domain → Direct delegation to one expert
├── Multi domain → Sequential/parallel multi-expert dispatch
├── Existing skill → Pass-through to matching skill
└── Ambiguous → Ask user for clarification
Expert Roster
| # | Expert | Command | Role |
|---|
| 1 | sub-team-lead | /sub-team-lead | Team orchestrator — classifies requests, routes to experts |
| 2 | sub-kopring-engineer | /sub-kopring-engineer | Kotlin/Java Spring Boot development (Hexagonal Architecture) |
| 3 | sub-test-engineer | /sub-test-engineer | Type-driven test generation (Java/Kotlin/TypeScript/Go) |
| 4 | sub-api-designer | /sub-api-designer | Contract-first API design (OpenAPI 3.1) |
| 5 | sub-code-reviewer | /sub-code-reviewer | Code review & refactoring (SOLID, code smells, tech debt) |
| 6 | sub-devops-engineer | /sub-devops-engineer | DevOps/CI-CD (Docker, GitHub Actions, K8s, Terraform) |
| 7 | sub-performance-engineer | /sub-performance-engineer | Performance analysis & optimization (JVM, DB, load testing) |
| 8 | engineering-workflow | /engineering-workflow | Architecture decisions across DB/BE/IF/SE domains |
| 9 | numerical | /numerical | Numerical computing verification (Python/Dart) |
| 10 | plugin-introspector | /plugin-introspector | Plugin monitoring & self-improvement |
| 11 | claude-autopilot | /claude-autopilot | Time-bounded autonomous task orchestration |
| 12 | sub-frontend-engineer | /sub-frontend-engineer | React/Vite frontend development (components, state, build) |
Pick Your Team
You don't need all 12 experts. Install only the ones that match your workflow.
Presets by Role
| If you are... | Recommended experts | Install |
|---|
| Backend developer (Kotlin/Java) | kopring + test + code-reviewer | 3 plugins |
| Backend + API design | kopring + test + api-designer + code-reviewer | 4 plugins |
| Full-stack team | team-lead + kopring + test + api-designer + code-reviewer + devops + frontend | 7 plugins |
| Performance-focused | kopring + test + performance | 3 plugins |
| Architecture/decision-making | engineering-workflow + code-reviewer | 2 plugins |
| Everything | All 12 | 12 plugins |
Standalone Experts (no dependencies)
Every expert works independently. There are no required dependencies between plugins.
# Example: Install only code-reviewer and test-engineer
/plugin install sub-code-reviewer@hungrytech-plugins
/plugin install sub-test-engineer@hungrytech-plugins
sub-team-lead is useful when you have 3+ experts installed — it auto-routes your requests. With 1-2 experts, just call them directly.
Sister-skill Integration (optional, not required)
When multiple experts are installed, they can hand off work to each other. This is automatic — no configuration needed. If a sister skill isn't installed, the expert simply skips the handoff and continues on its own.
sub-api-designer ──design──→ sub-kopring-engineer ──implement──→ sub-test-engineer
│ │
└── works fine alone works fine alone
sub-api-designer ──spec──→ sub-frontend-engineer ──UI──→ sub-devops-engineer
│ │ │
└── works fine alone └── works fine alone └── works fine alone
Install
Marketplace:
/plugin marketplace add hungrytech/hungrytech-claude-skills
/plugin install <skill-name>@hungrytech-plugins
Manual (single plugin):
git clone https://github.com/hungrytech/hungrytech-claude-skills.git
cp -r plugins/<skill-name>/skills/<skill-name> /path/to/project/.claude/skills/
Manual (multiple plugins):
git clone https://github.com/hungrytech/hungrytech-claude-skills.git
for skill in sub-kopring-engineer sub-test-engineer sub-code-reviewer; do
cp -r plugins/$skill/skills/$skill /path/to/project/.claude/skills/
done
Architecture
2-Layer Skill Structure
Every plugin shares the same core pattern:
- SKILL.md — Entry point that Claude Code reads. Core prompt + workflow definition.
- resources/ — On-demand loading per phase. Only injected when that phase starts (token efficiency).