← Back to Skills Marketplace
278
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install self-improving-agent-cn-skip
Description
AI自我改进与记忆系统 - 解决'同类错误反复犯、用户纠正不长记性'的痛点。自动捕获错误、用户纠正、最佳实践，并转化为长期记忆。
Usage Guidance
What to consider before installing:
- Expectation vs reality: the SKILL.md promises automatic capture/sync behavior, but the included scripts are simple CLI tools that only append/read JSONL files; there is no daemon, shell hook, or sync implementation. If you rely on "自动触发", ask the author how that is implemented or review any integration glue the agent will use.
- Sensitive data: memories may include command lines, errors, or user corrections (potentially secrets). The code stores data in plaintext under ~/.openclaw; consider whether you want that on disk, and enforce filtering/desensitization or encryption if needed.
- Behavior testing: run the scripts in a sandboxed account and inspect created files (errors.jsonl, corrections.jsonl, best_practices.jsonl) before enabling any agent automation that would call them.
- Integration risk: if you or your agent adds automation to call these scripts on every command failure, ensure the automation does not inadvertently capture tokens/passwords or automatically run corrective commands (the current code does not execute fixes, but an integration might).
- Ask the author for details about the claimed sync to project files and automatic triggers; absent that, treat the skill as a manual local memory store only.
Capability Analysis
Type: OpenClaw Skill
Name: self-improving-agent-cn-skip
Version: 1.0.0
The skill bundle provides a local logging system (using Python scripts like log_error.py and log_correction.py) to help an AI agent track errors and user preferences in ~/.openclaw/memory/. The code and SKILL.md instructions are consistent with the stated goal of creating a self-improving memory system, and there is no evidence of data exfiltration, malicious execution, or unauthorized access.
Capability Assessment
Purpose & Capability
The stated purpose is a local self-improvement/memory store for the agent, which matches the included scripts that append/read JSONL files under ~/.openclaw/memory/self-improving. However the SKILL.md repeatedly describes "自动触发" (automatic capture) and "自动同步" to project files (e.g., .learnings, AGENTS.md, MEMORY.md) that the provided scripts do not implement. The capability to automatically hook into shell commands or sync to other locations is claimed but not present in code.
Instruction Scope
Runtime instructions in SKILL.md instruct the agent to auto-record command failures, user corrections, and to check memories before executing commands. The shipped Python scripts only provide manual CLI entrypoints to append/read entries and do not include any listener/daemon, webhooks, or integration code that would automatically monitor commands or perform cross‑project writes. The doc's examples imply the agent will autonomously modify behavior (e.g., '自动改用 sudo'), but the scripts never execute or alter commands — they only write/read local files. This mismatch could lead to false expectations about what the skill will actually do; it also means an integrating agent would need additional glue that could introduce risk.
Install Mechanism
No install spec and no external downloads; the skill is instruction-plus-local Python scripts only. That minimizes install-time risk — nothing will be fetched from remote hosts during install.
Credentials
The skill requests no environment variables or credentials and only accesses files under the user's home (~/.openclaw/memory/self-improving). That access is proportionate to a local memory store. Note: memory files may contain sensitive information if the agent records raw command text, errors, or user corrections; there is no encryption or access control in the code.
Persistence & Privilege
The skill does not request always:true, does not change other skills' configs, and only writes to per-user paths. It persists data in user home but does not elevate privileges or force installation into all agents.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-improving-agent-cn-skip - After installation, invoke the skill by name or use
/self-improving-agent-cn-skip - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Self-Improving Agent 初始版本发布
- 自动记录AI错误、用户纠正、最佳实践，并转化为长期记忆，解决反复犯错与记忆丢失问题
- 支持命令执行前智能回顾相关“记忆”，减少重复错误
- 记忆自动跨项目同步：全局与项目级存储，支持写入AGENTS.md、MEMORY.md
- 检测并提示外部依赖变动、最佳实践更新、知识过时等场景
- 覆盖错误捕获、用户偏好纠正、做法改进全流程，自动触发，无需手动操作
Metadata
Frequently Asked Questions
What is Self Improving Agent Cn.Skip?
AI自我改进与记忆系统 - 解决'同类错误反复犯、用户纠正不长记性'的痛点。自动捕获错误、用户纠正、最佳实践，并转化为长期记忆。 It is an AI Agent Skill for Claude Code / OpenClaw, with 278 downloads so far.
How do I install Self Improving Agent Cn.Skip?
Run "/install self-improving-agent-cn-skip" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Self Improving Agent Cn.Skip free?
Yes, Self Improving Agent Cn.Skip is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Self Improving Agent Cn.Skip support?
Self Improving Agent Cn.Skip is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Self Improving Agent Cn.Skip?
It is built and maintained by Adin (@a-din); the current version is v1.0.0.
More Skills