An automated quality evaluation engine that audits AI agent skills against six critical dimensions and assigns performance badges.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install plugineval-core
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 plugineval-core 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 PluginEval Core?
PluginEval Core is a self-contained quality evaluation engine designed specifically for the Openclaw Skills ecosystem. It provides developers with a robust framework to measure the technical integrity and effectiveness of their AI agent skills without requiring external dependencies. By analyzing frontmatter, orchestration wiring, and token efficiency, it ensures that every skill meets the high standards required for reliable AI automation.
This engine is a cornerstone for developers who want to validate their Openclaw Skills before deployment or publication. It automates the detection of common pitfalls and provides a quantifiable score, helping maintain a high-quality codebase and a consistent user experience across different agent workflows.
PluginEval Core Use Cases
- Audit the quality of new AI skills prior to local installation or production deployment.
- Identify and resolve anti-patterns in existing Openclaw Skills to improve reliability and performance.
- Prepare skills for submission to public repositories by achieving Gold or Platinum status via automated scoring.
- Benchmark skill performance across specific dimensions like token efficiency and structural completeness.
How PluginEval Core Works
- The engine scans the target skill directory and parses the SKILL.md file for metadata, triggers, and structure.
- It performs Layer 1 static analysis to evaluate frontmatter quality and orchestration wiring effectiveness.
- An anti-pattern detection pass is run to identify architectural issues like bloated descriptions, missing triggers, or orphaned references.
- The tool calculates a weighted score across six specific quality dimensions defined in the quality framework.
- Finally, it generates a JSON report containing the numerical score, a letter grade, and a corresponding quality badge for the evaluated Openclaw Skills.
PluginEval Core Setup
Ensure you have Python 3 installed on your system. Since PluginEval Core is a self-contained skill, you can run it directly from your local Openclaw Skills directory.
# Navigate to the skill directory
cd ~/.openclaw/skills/plugineval-core
# Run a full evaluation on a target skill
python3 scripts/eval.py <path-to-your-skill-directory>
# Run only static analysis
python3 scripts/eval.py --layer1 <path-to-your-skill-directory>
PluginEval Core Data Schema & Taxonomy
The skill produces a structured JSON output representing the health and quality of Openclaw Skills.
| Attribute | Description |
|---|---|
skill |
The name or slug of the evaluated skill directory. |
score |
A numerical value (0-100) based on weighted dimensions. |
badge |
The quality tier assigned (Platinum, Gold, Silver, Bronze, or Needs Improvement). |
grade |
A letter grade (e.g., A+, B) representing the overall health. |
anti_patterns |
An array of detected issues and their associated scoring penalties. |
PluginEval Core Advanced Features
- Layer-specific analysis using flags like
--layer1or--layer2for focused architectural audits. - Granular anti-pattern detection to isolate specific weaknesses such as over-constrained prompts or bloated skill files.
- Automated badge assignment based on a rigorous weighted scoring logic that favors orchestration and frontmatter.
- Localized reference system that allows the engine to function entirely offline without external API dependencies.
SKILL.md
Loading
Related Openclaw Skills
Juejin Article Publisher
A professional publishing skill designed to help developers share technical articles and build personal brands on the Juejin community platform.
Xianyu Service Publisher
A professional framework for listing technical services and second-hand items on Alibaba's Xianyu platform to monetize skills effectively.
OSCHINA Technical Publisher
A professional automation tool for drafting, managing, and publishing technical articles on the OSCHINA open-source platform.
Search Agent
An intelligent AI research assistant that performs multi-source web searches and generates comprehensive, cited summaries.
Git Log Intelligence
A high-intelligence GitHub activity summarizer that filters noise and generates clean changelogs without cloning repositories.
Subhuti Coach Framework
A professional AI coaching framework integrating ICF Core Competencies with Buddhist-informed psychology for ethical, wise, and high-impact guidance.
Featured*
