Apollo Evolution is a framework for self-optimizing AI skills using biological principles of reproduction, mutation, and natural selection.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install apollo-evolution
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 apollo-evolution 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 Apollo Evolution?
Apollo Evolution introduces a biological paradigm to AI development, treating Openclaw Skills as evolving organisms rather than static code. By implementing a cycle of Reproduction, Mutation, and Selection, this skill allows your AI agents to discover the most effective prompts and parameters through practical iteration.
Instead of manual fine-tuning, Apollo Evolution automates the process of creating variants (mutations) of a skill, testing them against real-world inputs (environment), and retaining only the highest-performing versions (survival of the fittest). This ensures that the agent's logic remains dynamic and continuously improves its success rate over time.
Apollo Evolution Use Cases
- Iteratively improving system prompts to achieve higher accuracy in complex tasks.
- Finding the optimal balance of temperature and top_p settings for creative vs. technical outputs.
- Testing different structural workflows to find the most efficient step-by-step logic.
- Preserving successful logic patterns across different generations of Openclaw Skills development.
- Automating the A/B testing of AI agent behaviors in production environments.
How Apollo Evolution Works
- Copy: Select a base skill to optimize and create a master template version.
- Tweak: Introduce mutations by varying system prompts, adjusting LLM parameters, or modifying workflow steps.
- Test: Execute all variants using the same set of input data to ensure a fair comparison.
- Select: Evaluate performance scores and keep the best-performing version while discarding the others.
- Iterate: Use the winner as the parent for the next generation, repeating the process for continuous optimization.
Apollo Evolution Setup
To begin evolving your Openclaw Skills, use the provided bash scripts to manage the lifecycle of your agent variants:
# Step 1: Copy a skill to serve as the evolutionary starting point
bash scripts/apollo-evolution/copy.sh apollo-workflow
# Step 2: Generate mutated versions with different parameters
bash scripts/apollo-evolution/mutate.sh apollo-workflow-v1
# Step 3: Run automated tests and score the variants
bash scripts/apollo-evolution/test.sh apollo-workflow-v1 apollo-workflow-v2
# Step 4: Promote the best version to the next generation
bash scripts/apollo-evolution/select.sh
Apollo Evolution Data Schema & Taxonomy
Apollo Evolution maintains its state and history within a structured memory directory, ensuring transparency in the evolutionary process of your Openclaw Skills.
| Directory/File | Description |
|---|---|
.memory/evolution/{skill_name}/ |
Root directory for a specific skill's evolution data |
generations.json |
A record of each generation, tracking the lineage of improvements |
variants/ |
Folder containing the actual code and prompts for mutated versions (e.g., gen-001, gen-002) |
fitness.json |
Metadata containing the performance scores and evaluation metrics for each variant |
Apollo Evolution Advanced Features
- Prompt Mutation: Automatically toggles role definitions and output constraints to find peak performance.
- Parameter Mutation: Systematically scans temperature (0.5 to 1.0) and token limits to optimize cost and quality.
- Structural Mutation: Reconfigures task logic by adding validation steps or breaking down complex tasks.
- Epigenetic Learning: Passes successful experience and 'best practices' from previous generations to new Openclaw Skills.
- Multi-Agent Integration: Works alongside apollo-stem for self-updating and apollo-dream for memory consolidation.
SKILL.md
Loading
Related Openclaw Skills
Content Engine
A high-performance content automation skill that transforms single briefs into 30+ platform-optimized assets with consistent brand voice.
Auto Design
An intelligent design system selector that automatically identifies and applies the best visual styles for your project based on context and intent.
Cold Email Sequence
A specialized AI skill for generating high-converting, multi-stage cold email sequences with built-in deliverability optimization and A/B testing.
Design Extractor
A powerful utility that scans your source code to automatically generate comprehensive design system documentation and DESIGN.md files.
ThinkPHP UI Restoration
A specialized agent skill for generating and restoring ThinkPHP UI components while maintaining project-specific style conventions and template logic.
Explore Europe
A specialized travel orchestration skill for AI agents to plan European itineraries with real-time data from Fliggy.
Featured*
