EvoAgentX Workflow

Bridge EvoAgentX (1000+ star open-source framework) with OpenClaw. Enables self-evolving agentic workflows - workflows that automatically improve over time t...

Security Scan

The skill's files and instructions are consistent with an EvoAgentX integration and do not request extra credentials or suspicious privileges, but you should still audit the third-party Python package before installing it into a real environment.

Purpose & Capability

Name, description, SKILL.md, and the provided CLI script all align: this is an integration wrapper for the EvoAgentX framework. Requiring python3/pip and suggesting 'pip install evoagentx' is coherent with the stated purpose.

Instruction Scope

Runtime instructions only guide installing EvoAgentX, creating/editing workflows, and running the included CLI. The script reads/writes a workflow file in the current directory and checks for optional 'openai' integration, but it does not read unrelated system files or request secrets.

Install Mechanism

The skill is instruction-only but the SKILL.md metadata suggests installing the 'evoagentx' package via pip or cloning from GitHub — expected for this purpose. Note: pip installations (and git-installed packages) execute package installation code (setup scripts), so review the upstream package source or install in an isolated environment (venv/container). Also there is a minor metadata inconsistency: registry shows no separate install spec while SKILL.md contains an install entry.

Credentials

No environment variables, credentials, or config paths are required. The script does not request tokens or secrets; it only optionally checks for the presence of 'openai' to report an available integration.

Persistence & Privilege

always:false and user-invocable:true (normal). The skill writes workflow files into the current working directory when asked (create-workflow) but does not modify other skills or system-wide agent configuration.

Assessment

This skill appears coherent and low-risk for its intended purpose, but take these precautions before installing: (1) review the 'evoagentx' PyPI/GitHub source yourself (or install inside a virtualenv or container) because pip installs can run arbitrary code at install time; (2) inspect any generated workflow files before running them; (3) avoid installing as root and keep the package isolated if you plan to test; (4) if you rely on the SKILL.md claim that 'all evolution happens locally', verify the EvoAgentX package and any integrations you enable (e.g., OpenAI) to ensure they don't send data externally.

EvoAgentX Workflow Integration

Integrates the EvoAgentX framework with OpenClaw for self-evolving agentic workflows.

When to Use This Skill

Use this skill when:

• Building multi-agent workflows that need to evolve over time
• Evaluating and optimizing existing agent workflows
• Implementing the Genome Evolution Protocol (GEP)
• Creating self-improving agent systems
• Migrating static workflows to self-evolving ones

Quick Start

CLI Usage

This skill provides a command-line interface for EvoAgentX operations:

# Check if EvoAgentX is installed
python3 scripts/evoagentx_cli.py status

# Get installation instructions
python3 scripts/evoagentx_cli.py install

# Show usage examples
python3 scripts/evoagentx_cli.py examples

# Create a workflow template
python3 scripts/evoagentx_cli.py create-workflow \
  --name ResearchWorkflow \
  --description "A research automation workflow"

# Check EvoAgentX status
python3 scripts/evoagentx_cli.py check

Installation

# Install EvoAgentX framework
pip install evoagentx

# Verify installation
python3 -c "import evoagentx; print(evoagentx.__version__)"

1. Create a Basic Workflow

After running create-workflow, edit the generated Python file:

from evoagentx import Agent, Workflow

class MyWorkflow(Workflow):
    async def execute(self, context):
        # Your workflow logic here
        result = await self.run_agents(context)
        return result

2. Enable Self-Evolution

from evoagentx.evolution import EvolutionEngine

engine = EvolutionEngine()
optimized_workflow = await engine.evolve(
    workflow=MyWorkflow(),
    iterations=10,
    evaluation_criteria={"accuracy": 0.95}
)

Core Concepts

Workflows

• Multi-agent orchestration
• State management
• Tool integration

Evolution Strategies

• TextGrad: Prompt optimization
• AFlow: Workflow structure optimization
• MIPRO: Multi-step reasoning optimization

Genomes

Encoded success patterns containing:

• Task type classification
• Approach methodology
• Outcome metrics
• Context requirements

Common Patterns

Pattern 1: Research Workflow Evolution

# Start with basic research workflow
workflow = ResearchWorkflow()

# Evolve for better results
evolution = await workflow.evolve(
    dataset=research_queries,
    metric="comprehensiveness"
)

Pattern 2: Tool Selection Optimization

# EvoAgentX automatically selects optimal tools
workflow = AgentWorkflow(
    tools=["web_search", "browser", "file_io"],
    auto_select=True
)

Security Considerations

• All evolution happens locally (no data exfiltration)
• Genomes contain no credentials
• Evaluation uses synthetic data when possible

References

• EvoAgentX GitHub: https://github.com/EvoAgentX/EvoAgentX
• Documentation: https://evoagentx.github.io/EvoAgentX/
• arXiv Paper: https://arxiv.org/abs/2507.03616

Version

1.0.0 - Initial release with core EvoAgentX integration