Skill Evolution for Openclaw

A powerful meta-skill designed to guide other AI agent skills through interactive directional evolution and cognitive phagocytosis.

pillar-wang
v1.0.1
Jun 7, 2026
0
616
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install skill-evolution

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install skill-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 Skill Evolution?

The skill-evolution framework operates as an advanced Meta-Skill tailored for AI agent runtime optimizations. Instead of relying purely on autonomous hill-climbing algorithms, it introduces human-in-the-loop, surgical iterations for agent capabilities. By executing either targeted refinement or absorbing external mechanics, it acts as an evolutionary catalyst for system prompts and workflow configurations.

This infrastructure ensures complete execution reliability across various environments like WorkBuddy, Codex, and Clawdbot. By natively handling byte-level matching safeguards and granular file tracking, it systematically upgrades your Openclaw Skills inventory while ensuring absolute runtime stability via precise user-approved modifications.

Skill Evolution Use Cases

  • When an AI agent skill requires precise updates based on runtime edge cases or domain shifts.
  • When you need an agent to absorb logic, reference materials, or tools from an external document (PDF, Word, Markdown, TXT).
  • When you want to extract and codify patterns, corrections, and user preferences directly from an active conversation history into a permanent skill structure.
  • When coordinating multi-agent capabilities by synthesizing cross-skill logic without duplicating systemic friction.

How Skill Evolution Works

  1. Target Identification: The framework parses the local environment using glob indexing to locate and validate the target SKILL.md template file across relative deployment folders.
  2. Path Selection: The user designates whether the target should undergo Path A (Directional Evolution) or Path B (Phagocytosis/Fusion Execution).
  3. Automated Guardrail Check: A multi-layered atomic backup sequence clones the existing configuration state; any operational failure immediately halts executions.
  4. Mutation & Code Diff Synthesis: The platform isolates weaknesses based on structural archetypes, crafts an atomic refinement scheme, and formats a visual diff preview.
  5. Sandbox Verification: The agent introduces suggested variations into the target runtime and expects immediate human feedback or real-world execution testing.
  6. Retention or Rollback: Upon validation, changes commit alongside systematic log generations; failures trigger a byte-level revert loop to restore initial parameters.

Skill Evolution Setup

Prerequisites

Ensure your local environment includes standard markdown processing capabilities and python utilities if parsing complex documents.

# Standard deployment within your agent framework directory
mkdir -p ./skills/skill-evolution

# If swallowing external assets, ensure parsing libraries are ready
pip install python-docx pdfplumber pymupdf

Configuration

Verify that your references/evolution-directions.md and references/resources.md configurations populate properly before invoking commands to evolve your Openclaw Skills portfolio.

Skill Evolution Data Schema & Taxonomy

The meta-skill handles backup data, transformation workflows, and execution tracking within a dedicated .skill-backups taxonomy.

Filepath Type Functional Role
.skill-backups/{skill-name}/evolution.log Log File Tracks step-by-step directional mutations, logic modifications, and outcomes.
.skill-backups/{skill-name}/phagocytosis.log Log File Contains explicit ingestion tracking, asset origins, and applied filters.
.skill-backups/{skill-name}/SKILL.md.{timestamp} Backup Raw byte-level clone of the source markdown file prior to mutations.
.skill-backups/{skill-name}/SKILL.md.{timestamp}.post Stage Replica Staged evolution output retained for safety comparisons before commit.

Skill Evolution Advanced Features

  • Three-Layer Rollback Shield: Executes a progressive recovery routine starting with file-level byte checks, falling back to Git stash extractions, and escalating to manual intervention freezes upon structural anomalies.
  • Multi-Format Ingestion Processor: Dynamically splits, parses, and strips conflicting frontmatter tags out of incoming .docx, .pdf, .md, and .txt files.
  • Synergistic Darwin-Skill Coupling: Pairs seamlessly with automated grading loops; darwin-skill isolates low-performing criteria, and skill-evolution applies targeted human-guided remedies.
  • Cognitive Phagocytosis Engine: Extracts reusable logical boundaries, workflow chains, system boundaries, and structural tables while safely discarding conflicting runtime tokens up to a strict 10-item limit per batch.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*