← Back to Skills Marketplace
440
Downloads
0
Stars
2
Active Installs
2
Versions
Install in OpenClaw
/install self-improving-agent-next
Description
记录错误、纠正、能力缺口与最佳实践，形成可复用的持续改进闭环。适用于：命令失败、用户纠正、外部 API/工具异常、发现更优做法、提出新能力需求，以及任务前复盘历史经验。
Usage Guidance
This skill appears to do what it says: remind the agent to record learnings and provide safe helper scripts. Before enabling it, consider: (1) whether you want hooks enabled at the user-global level (~/.claude) — enabling globally runs the activator/error-detector on every session, (2) review and, if desired, sandbox or audit the scripts (scripts/extract-skill.sh writes files under the current workspace), (3) if you don't want automatic PostToolUse checks, only enable the activator or add matcher filters per the docs, and (4) ensure .learnings/ file locations and permissions are acceptable for your workflow. Overall the package is internally consistent and proportional to its stated purpose.
Capability Analysis
Type: OpenClaw Skill
Name: self-improving-agent-next
Version: 2.0.1
The 'self-improving-agent' bundle is a legitimate tool designed to help AI agents maintain a structured log of errors, user corrections, and best practices within a workspace. It includes utility scripts like 'extract-skill.sh' for scaffolding new skills and 'error-detector.sh' for identifying command failures, along with OpenClaw hooks to inject reminders into the agent's context. The implementation is transparent, well-documented, and includes safety checks in its shell scripts to prevent directory traversal during file creation.
Capability Assessment
Purpose & Capability
Name/description (capture errors/learnings/promote to skills) match the included files: templates, hook handlers, activator/error-detector scripts, and an extractor for promoting learnings to skills. There are no unrelated environment variables, binaries, or external services requested.
Instruction Scope
SKILL.md instructs writing structured records under .learnings/ and provides templates and promotion rules; the scripts and hook simply surface reminders or detect command output errors. There are no instructions to read unrelated system files, exfiltrate data, or call unexpected external endpoints. The docs reference OpenClaw session utilities (sessions_send, sessions_spawn) as available platform functions but do not instruct the skill to automatically use them.
Install Mechanism
No install spec; this is instruction-and-script based. All code is packaged in the skill and scripts operate locally. No downloads from external URLs or archive extraction are present.
Credentials
The skill declares no required env vars or credentials. The error-detector script reads a CLAUDE_TOOL_OUTPUT environment variable (documented as part of the platform), which is proportional to detecting tool failures. No secrets or unrelated credentials are requested.
Persistence & Privilege
always is false and the skill does not request persistent global privileges. Hook files can be enabled by the user (project or user-level) and scripts run with the agent's normal runtime permissions — this is expected. The extractor writes files under a relative skills directory and includes checks to avoid absolute paths or .. segments.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-improving-agent-next - After installation, invoke the skill by name or use
/self-improving-agent-next - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.1
Rewrite SKILL.md fully in Chinese
v2.0.0
Initial release
Metadata
Frequently Asked Questions
What is self-improving-agent?
记录错误、纠正、能力缺口与最佳实践，形成可复用的持续改进闭环。适用于：命令失败、用户纠正、外部 API/工具异常、发现更优做法、提出新能力需求，以及任务前复盘历史经验。 It is an AI Agent Skill for Claude Code / OpenClaw, with 440 downloads so far.
How do I install self-improving-agent?
Run "/install self-improving-agent-next" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is self-improving-agent free?
Yes, self-improving-agent is completely free (open-source). You can download, install and use it at no cost.
Which platforms does self-improving-agent support?
self-improving-agent is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created self-improving-agent?
It is built and maintained by 微云字节 (@weiyunbyte); the current version is v2.0.1.
More Skills