← Back to Skills Marketplace
316
Downloads
2
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install self-improvement-system
Description
Runs a continuous self-improvement loop that helps the agent learn from mistakes, extract lessons, and refine its behaviour over time. Use when the user says...
Usage Guidance
This skill appears coherent and does what it says: it will read and write persistent files in the agent workspace to log mistakes, extract lessons, and run audits. Before installing, consider: (1) Review any existing mistakes.md / lessons.md / soul.md files so they don’t already contain sensitive user data. (2) Ensure your agent environment enforces the platform's privacy boundaries — the skill relies on the agent obeying its own rules about not logging PII or user-provided content. (3) Be aware the skill will persist state across sessions (archiving and rotation rules are built in). If you need stricter guarantees, periodically inspect or back up the created files and consider limiting the agent's filesystem or network access to reduce accidental data exposure.
Capability Analysis
Type: OpenClaw Skill
Name: self-improvement-system
Version: 1.2.0
The skill implements a structured self-improvement and logging system designed to help the agent learn from reasoning errors across sessions. It manages local files such as 'soul.md' and 'lessons.md' to store abstract behavioral principles and workflows, and it includes explicit, detailed privacy instructions in both SKILL.md and references/protocol.md to prevent the logging of PII, credentials, or sensitive user data.
Capability Assessment
Purpose & Capability
Name and description match the runtime instructions: the skill maintains per-agent files (mistakes.md, lessons.md, soul.md, playbook.md, session-log.md, archive/*) and runs audits and pattern detection. It does not request unrelated binaries, credentials, or network access.
Instruction Scope
All instructions stay within the stated purpose (logging, lesson extraction, audits, playbooks). They require reading and writing files in the agent workspace and instruct the agent to create missing files. The privacy rules are explicit (no user data, no verbatim quotes). The main caveat is behavioral: the skill gives the agent judgment calls (paraphrase vs omit) which, if misapplied by the agent, could lead to unsafe logs — this is a usability/safety risk but not an incoherence.
Install Mechanism
Instruction-only skill with no install spec or code files. This is the lowest-risk install pattern and matches the described functionality.
Credentials
The skill requests no environment variables, credentials, or special config paths. It relies only on reading/writing local files described in the SKILL.md, which is proportionate for a local self-improvement system.
Persistence & Privilege
The skill persists state across sessions by creating and updating files (mistakes.md, lessons.md, soul.md, playbook.md, session-log.md, archive). always is false (not force-included), but it is intended to trigger at session start and periodically. Persistent storage and autonomous invocation are coherent with its purpose; users should be aware these files are retained between runs and could contain agent-generated summaries of past interactions.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-improvement-system - After installation, invoke the skill by name or use
/self-improvement-system - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.2.0
**This version introduces strict privacy safeguards to all self-improvement logging and process files.**
- Added detailed privacy and data safety section: only abstract descriptions of process errors may be logged, never user data.
- Banned logging of any personally identifiable information, user-provided content, verbatim user messages, credentials, or sensitive details.
- Required paraphrasing or omission of any mistake description that could expose user information.
- Applied privacy rules to all core files, including session-log entries.
- Updated mindset and process to prioritize privacy over volume in logging and lesson extraction.
v1.1.0
**Self-improvement Skill v1.1.0 Changelog**
- Major simplification and overhaul: moved from detailed file-based workflow to a focused self-improvement loop and session protocols.
- Removed legacy and OpenClaw-specific supporting files, log templates, and hook references.
- Introduced core files: mistakes.md, lessons.md, soul.md, playbook.md, session-log.md, and archive directory.
- New guidance for startup, mistake logging, and session close; mandates pre-action review of behavioural files.
- Added clear criteria for when to log mistakes and how to rotate/archive logs.
- Reference protocol and format details moved to a single concise file.
Metadata
Frequently Asked Questions
What is Self-Improvement System?
Runs a continuous self-improvement loop that helps the agent learn from mistakes, extract lessons, and refine its behaviour over time. Use when the user says... It is an AI Agent Skill for Claude Code / OpenClaw, with 316 downloads so far.
How do I install Self-Improvement System?
Run "/install self-improvement-system" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Self-Improvement System free?
Yes, Self-Improvement System is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Self-Improvement System support?
Self-Improvement System is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Self-Improvement System?
It is built and maintained by assafster (@assafster); the current version is v1.2.0.
More Skills