An advanced multi-agent framework for designing, auditing, and maintaining high-quality Openclaw Skills through automated design and testing loops.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skill-engineer
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install skill-engineer using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Skill Engineer?
The Skill Engineer is a professional-grade toolkit designed to manage the full lifecycle of agent capabilities within your environment. Built on the core principle that builders should not evaluate their own work, this skill implements a sophisticated multi-agent architecture. It orchestrates independent Designer, Reviewer, and Tester subagents to ensure every skill produced meets rigorous technical standards. By utilizing this system, developers can automate the creation of both Capability Uplift and Encoded Preference skills, ensuring that their Openclaw Skills are robust, version-compatible, and aligned with organizational processes. This workflow significantly reduces manual effort while increasing the reliability of AI-driven automation.
At its heart, the Skill Engineer provides a structured pathway for transforming natural language requirements into validated technical artifacts. It bridges the gap between raw model capabilities and specialized workflows, making it an essential component for teams looking to scale their agent infrastructure. Whether you are building complex data pipelines or simple task automations, this skill ensures your Openclaw Skills are optimized for performance and security from the first iteration.
Skill Engineer Use Cases
- Designing new agent capabilities from scratch using natural language requirements.
- Auditing existing Openclaw Skills to identify logic gaps, security risks, or outdated API usage.
- Refactoring legacy skills to align with the latest system specifications and frontmatter standards.
- Implementing automated testing and evaluation frameworks (evals) for complex agent workflows.
- Performing periodic kit-wide audits to ensure naming and functional consistency across multiple agents.
How Skill Engineer Works
- Requirement Gathering: The Director agent collects problem descriptions, target audiences, and interaction constraints from the user context.
- Context Grounding: The system queries DeepWiki and vector memory to ensure all design decisions are based on the latest API behavior and historical session data.
- Multi-Agent Design: A Designer subagent generates the initial SKILL.md, skill.yml, and associated test scripts in an isolated session.
- Independent Review: A Reviewer subagent evaluates the artifacts against a 33-point quality rubric, providing a structured score and identifying potential issues.
- Empirical Testing: A Tester subagent performs self-play validation to verify trigger accuracy and functional integrity under realistic scenarios.
- Iterative Refinement: If quality gates (score < 28/33) are not met, the Orchestrator routes feedback back to the Designer for up to three refinement loops.
- Final Delivery: Once passed, the skill is finalized with a quality scorecard and committed to the repository for deployment.
Skill Engineer Setup
Ensure your environment is ready for advanced Openclaw Skills development by verifying mandatory dependencies. First, install the DeepWiki skill to ground your designs in actual source code:
# Install deepwiki from the official repository
openclaw install liaosvcaf/openclaw-skill-deepwiki
Next, enable vector memory in your openclaw.json configuration to allow the agent to access historical design decisions:
{
"memory": {
"enabled": true
}
}
For complex, multi-phase workflows, configure your agent to allow nested subagents by setting the spawn depth:
{
"agents": {
"defaults": {
"subagents": {
"maxSpawnDepth": 2
}
}
}
}
Skill Engineer Data Schema & Taxonomy
The Skill Engineer organizes Openclaw Skills into a standardized directory structure to ensure portability and clarity across different agent kits.
| File | Description | Requirement |
|---|---|---|
| SKILL.md | Primary file containing frontmatter, descriptions, and instructions. | Mandatory |
| skill.yml | Metadata for publishing to the ClawHub registry and managing versions. | Optional |
| README.md | Human-readable documentation including the final Quality Scorecard. | Recommended |
| tests/ | Contains evals.json, benchmark results, and regression test cases. | Recommended |
| scripts/ | Deterministic helper scripts for specialized validation or CLI logic. | Optional |
| references/ | Deep-dive documentation or research notes linked from the main skill. | Optional |
Skill Engineer Advanced Features
- Orchestrator Subagent Pattern: Utilizes a dedicated supervisor subagent to prevent pipeline stalls during long-running design cycles.
- Capability vs. Preference Taxonomy: Tailors testing strategies based on whether the skill adds new model abilities or enforces specific organizational processes.
- Evals Framework: Supports A/B blind testing and automated regression evals to track the performance of Openclaw Skills over time.
- DeepWiki Staleness Protocol: Automatically verifies version tags to prevent the use of deprecated APIs or internal mechanics.
- Automatic Quality Scorecards: Generates detailed markdown tables in README files to provide a quality certificate for every shipped skill.
- Skill Retirement Protocol: Identifies when base model improvements have made specific capability uplift skills redundant.
SKILL.md
Loading
Related Openclaw Skills
Save to Obsidian
A specialized automation skill for saving markdown content and diagrams from an AI agent directly into a remote Obsidian vault via secure SSH transfer.
Rotate OpenRouter Key
A specialized utility to securely replace OpenRouter API keys across multiple configuration files and environment variables.
ROSE Docker Build
A specialized environment for building the ROSE source-to-source compiler using Docker to ensure toolchain compatibility and dependency isolation.
ROSE Container Tools
A streamlined environment for building and running ROSE-based source code analysis tools within a Docker container to ensure toolchain compatibility.
Mermaid Diagram Generator
A powerful automation skill for creating professional, consistently styled Mermaid diagrams from technical content using smart templates.
WeChat Article Extractor
A specialized tool for converting WeChat public account articles into clean, archived Markdown files with full image and metadata support.
METADATA
Featured*
