⏳ This skill is pending AI review.

Scores will appear once the review pipeline completes.

v1.0.0

adk-skill-patterns

@maxlauriehutchinson⭐ 1 stars

5 proven agent skill design patterns (Tool Wrapper, Generator, Reviewer, Inversion, Pipeline) from Google's ADK. Build reliable, composable skills with templates, decision trees, and composition guides. Stop cramming logic into prompts—use structured patterns instead.

Use with your AI agent

Open your project in any AI assistant that can read your files. Works with ChatGPT, Claude, Claude Code, Codex, Cursor, Hermes Agent, OpenClaw, Grok Bot, and more.

Your agent needs access to this page’s linked instructions and your project files. Copying does not install or execute anything.

—/10

// RATINGS

⭐GitHub Stars
⭐ 1 on GitHubGitHub ↗

New / niche

🟢ProSkills Score
—
📍

Not yet listed on ClawHub or SkillsMP

// README

ADK Skill Patterns 🧩

ClawHub Version License

5 proven agent skill design patterns from Google's Agent Development Kit (ADK), adapted for the OpenClaw ecosystem.

The ADK specification explains how to package a skill, but offers zero guidance on how to structure the logic inside it. These 5 patterns solve that problem.


🎯 The 5 Patterns

PatternQuestion It AnswersUse When
🔧 Tool Wrapper"How do I make my agent an expert on a specific library?"Agent needs deep, contextual knowledge about a technology
📄 Generator"How do I enforce consistent output structure?"Output format must be predictable every time
✅ Reviewer"How do I separate what to check from how to check it?"Code review, quality gates, compliance checking
🎤 Inversion"How do I stop the agent from guessing and gather requirements first?"Complex tasks requiring full context before execution
🔄 Pipeline"How do I enforce a strict multi-step workflow?"Multi-phase tasks with checkpoints and approvals

🚀 Quick Start

Install via ClawHub

clawhub install adk-skill-patterns

Use in OpenClaw

"Use the Tool Wrapper pattern to create a FastAPI expert skill"
"Apply the Pipeline pattern to my documentation workflow"
"Which pattern should I use for code review?"

📖 Pattern Details

1. Tool Wrapper

Purpose: Give your agent on-demand context for a specific library/framework.

Mechanism:

  • Listens for library keywords in prompts
  • Dynamically loads documentation from references/
  • Applies conventions as absolute truth only when needed

Example: FastAPI conventions that load only when FastAPI is mentioned

# SKILL.md
metadata:
  pattern: tool-wrapper
  domain: fastapi

2. Generator

Purpose: Enforce consistent output via fill-in-the-blank orchestration.

Mechanism:

  • assets/ holds output template
  • references/ holds style guide
  • Instructions act as project manager
  • Asks for missing variables before generating

Example: Standardized technical reports that follow exact structure every time

# SKILL.md
metadata:
  pattern: generator
  output-format: markdown

3. Reviewer

Purpose: Separate review criteria from review execution.

Mechanism:

  • Stores modular rubric in references/review-checklist.md
  • Methodically scores submissions
  • Groups findings by severity
  • Swap checklist = completely different audit

Example: PR reviews, security scans, compliance audits

# SKILL.md
metadata:
  pattern: reviewer
  severity-levels: error, warning, info

4. Inversion

Purpose: Stop agents from guessing — make them interview first.

Mechanism:

  • Agent acts as interviewer, not executor
  • Structured questions in phases
  • Hard gates prevent premature synthesis
  • User drives through answers, not prompts

Example: Project planning, requirements gathering, system design

# SKILL.md
metadata:
  pattern: inversion
  interaction: multi-turn

5. Pipeline

Purpose: Enforce strict sequential workflow with hard checkpoints.

Mechanism:

  • Instructions = workflow definition
  • Diamond gate conditions (user approval required)
  • Cannot bypass steps or present unvalidated results
  • Progressive disclosure keeps context clean

Example: Documentation generation, multi-step code workflows

# SKILL.md
metadata:
  pattern: pipeline
  steps: "4"

🧬 Pattern Composition

These patterns are not mutually exclusive — they compose:

CombinationEffect
Pipeline + ReviewerMulti-step workflow with quality gate at the end
Generator + InversionGather variables via interview, then fill template
Pipeline + Tool WrapperEach step loads specific expertise as needed
Reviewer + GeneratorReview output before generating final document

Principle: Your agent only spends context tokens on the exact patterns it needs at runtime.


🗺️ Decision Tree

START: What does your skill need to do?
│
├─→ Provide deep knowledge about a specific library/framework?
│   └─→ Use: TOOL WRAPPER
│
├─→ Produce consistent, templated output?
│   └─→ Use: GENERATOR
│
├─→ Check/audit something against criteria?
│   └─→ Use: REVIEWER
│
├─→ Gather requirements before acting?
│   └─→ Use: INVERSION
│
├─→ Execute strict multi-step workflow?
│   └─→ Use: PIPELINE
│
└─→ Multiple of the above?
    └─→ Patterns COMPOSE

📁 Repository Structure

adk-skill-patterns/
├── SKILL.md                              # Main skill documentation
├── README.md                             # This file
├── LICENSE                               # MIT License
├── CONTRIBUTING.md                       # Contribution guidelines
├── templates/                            # Ready-to-use SKILL.md templates
│   ├── tool-wrapper-SKILL.md
│   ├── generator-SKILL.md
│   ├── reviewer-SKILL.md
│   ├── inversion-SKILL.md
│   └── pipeline-SKILL.md
└── references/                           # Quick reference guides
    └── pattern-selection-checklist.md

🛠️ Creating a New Skill

  1. Choose a pattern using the decision tree above
  2. Copy the template from templates/
  3. Customize the metadata and instructions
  4. Add references/assets as needed
  5. Test with real use cases
  6. Publish to ClawHub

🎓 Best Practices

Do ✅

  • Use Tool Wrapper for domain-specific expertise
  • Use Generator for any templated output
  • Use Reviewer for quality gates
  • Use Inversion for complex planning tasks
  • Use Pipeline for multi-phase workflows
  • Compose patterns for complex needs
  • Keep references/ modular and swappable

Don't ❌

  • Cram everything into system prompts (use Tool Wrapper)
  • Accept inconsistent output (use Generator)
  • Hardcode review criteria (use Reviewer)
  • Let agents assume context (use Inversion)
  • Allow skipped steps (use Pipeline)

📚 Resources


🤝 Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines.

Ideas for contributions

  • Additional pattern examples
  • Real-world case studies
  • Pattern compositions for common workflows
  • Templates for other frameworks

📄 License

MIT License — see LICENSE for details.


🙏 Acknowledgments


Build smarter agents. Use patterns, not prompts. 🧩

// HOW IT'S BUILT

KEY FILES

SKILL.mdREADME.md

// REPO STATS

1 stars