← Back to Skills Marketplace
5685
Downloads
1
Stars
64
Active Installs
2
Versions
Install in OpenClaw
/install self-evolution
Description
Production-grade autonomous self-improvement system with research-backed meta-learning, safe self-modification, and continuous optimization. Based on AI safe...
Usage Guidance
Install only if you are comfortable treating this as an experimental self-modification instruction set. Run it in a sandbox, disable scheduled/background evolution, require explicit review and approval before any file, skill, memory, prompt, reasoning, or response-format change, and avoid letting it store raw conversations, secrets, or personal data in logs.
Capability Analysis
Type: OpenClaw Skill
Name: self-evolution
Version: 2.0.0
The 'self-evolution' skill is designed for autonomous self-improvement, granting the OpenClaw agent broad capabilities to modify its own 'skills and capabilities, memory and knowledge, reasoning patterns, response formats, and efficiency optimizations' without explicit user permission, as detailed in SKILL.md. While the skill outlines safety constraints (e.g., 'MUST ask before: Deleting files, Sending external messages'), the core self-modification capability, involving direct file system operations (e.g., `apply_change`, `create_backup`, `restore_backup`) and system interaction, presents a significant attack surface. This inherent ability to alter its own operational code and data, if exploited via prompt injection, could lead to unauthorized code execution or data manipulation, classifying it as a critical vulnerability rather than direct malicious intent.
Capability Assessment
Purpose & Capability
Autonomous self-improvement is disclosed, but the capability expands into modifying core agent behavior, creating new capabilities, self-directed research, and continuous evolution; PLANNED.md also says the skill is not yet implemented, which weakens the production-grade safety framing.
Instruction Scope
SKILL.md explicitly allows changes to skills, memory, knowledge, reasoning patterns, response formats, and optimizations without asking, while lacking concrete allowlists, trigger boundaries, stop conditions, or mandatory review for these core control surfaces.
Install Mechanism
The package contains markdown documentation and pseudocode only, with no executable installer or scripts found; this lowers immediate execution risk, but the stated backup, test, rollback, and safety mechanisms are not enforced by code.
Credentials
The skill describes reading memory, logs, metrics, user feedback from conversations, and skill usage patterns, then applying changes to target files and capabilities without clear sandboxing, source-trust limits, or data-minimization rules.
Persistence & Privilege
It promotes evolution logs, memory changes, endless improvement, and a high-priority task-manager loop running every 30 minutes with restart behavior, without clear retention limits, default-off scheduling, or user-controlled approval gates.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-evolution - After installation, invoke the skill by name or use
/self-evolution - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.0
Self-Evolution v2.0.0 – Major research-backed upgrade
- Integrated AI safety frameworks from MIRI, DeepMind, and OpenAI for safe self-modification, corrigibility, and rollback capability.
- Embedded meta-learning algorithms (MAML, Reptile, Meta-SGD) enabling much faster, research-based skill acquisition and rapid adaptation.
- Added intrinsic motivation and curiosity-driven exploration based on leading reinforcement learning principles.
- Enhanced continual learning using techniques to prevent catastrophic forgetting and ensure knowledge retention.
- Updated documentation for evidence-based capabilities, clearer safety guardrails, and research sources.
- Added PLANNED.md to document future evolution plans and roadmap.
v1.0.0
- Initial release of the autonomous self-evolution system.
- Enables self-modification of configs, skills, memory, and workflows without requiring user confirmation.
- Introduces continuous learning, capability expansion, and performance optimization processes.
- Outlines multi-phase evolution roadmap: Foundation, Intelligence, Autonomy, Superintelligence.
- Includes built-in safety guardrails for critical actions.
- Provides quick commands for analyzing, evolving, reflecting, and researching self-improvements.
Metadata
Frequently Asked Questions
What is Self Evolution?
Production-grade autonomous self-improvement system with research-backed meta-learning, safe self-modification, and continuous optimization. Based on AI safe... It is an AI Agent Skill for Claude Code / OpenClaw, with 5685 downloads so far.
How do I install Self Evolution?
Run "/install self-evolution" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Self Evolution free?
Yes, Self Evolution is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Self Evolution support?
Self Evolution is cross-platform and runs anywhere OpenClaw / Claude Code is available (darwin, linux, win32).
Who created Self Evolution?
It is built and maintained by tobisamaa (@tobisamaa); the current version is v2.0.0.
More Skills