Self-Improving Agent
v0.2.0基于对经验的持续学习，不断优化 Agent 能力。适用于完成重要任务后、出现错误时、会话结束时，或用户输入“自我进化”“总结经验”“从经验中学习”等指令时触发。
⭐ 0· 534·0 current·0 all-time
bydry3@initail
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name and description promise a self-improvement loop that extracts patterns, stores episodic/semantic/working memories, proposes skill updates, and requests user confirmation before applying changes. The SKILL.md and accompanying templates implement those behaviors (memory folders, patterns.json, templates for corrections/patterns/validation). There are no unexpected credentials, binaries, or external hosts required that would contradict the declared purpose.
Instruction Scope
Instructions require the agent to discover the workspace root and to read workspace memory files (MEMORY.md and memory/YYYY-MM-DD.md) 'Always' to gather context. They also allow appending to those files and performing full CRUD inside memory/self-improving/, but state that any skill-file modifications require explicit user confirmation. Reading and modifying files inside the workspace is coherent for a self-improving agent, but the instruction to 'always' read memory files is broad and could surface any sensitive content the user stores in the workspace. The skill does not instruct exfiltration or contact with external endpoints.
Install Mechanism
This is an instruction-only skill with no install spec and no code files to execute. That minimizes installation risk — nothing is downloaded or written to disk by an installer.
Credentials
The skill declares no required environment variables, no credentials, and no config paths. Its runtime behavior is limited to reading and writing files inside the workspace, which aligns with its stated purpose. There are no requests for unrelated secrets or external service tokens.
Persistence & Privilege
The skill is not marked always:true and is user-invocable. It may be invoked autonomously by agents (default platform behavior), but the skill documents a User Confirmation Gate: all skill-file modifications require explicit user approval before applying. It does propose skill updates and writes to agent memory directories (within workspace), which is consistent with a self-improvement feature when user confirmation is required.
Assessment
This skill appears coherent with its purpose, but exercise caution before approving any changes it proposes. Practical steps:
- Review proposed skill modifications and memory appends carefully (diffs) before approving.
- Back up MEMORY.md and other workspace files so you can revert unwanted changes.
- Avoid storing secrets or sensitive data in workspace files the agent will read (MEMORY.md, other project files) or move such secrets out of the workspace.
- Consider testing the skill in an isolated or dummy workspace first so you can observe its proposals without risk to production data.
- If the skill proposes code or skill-file changes, require a manual code review or run tests before applying them.
- If you need stricter guarantees, prefer a read-only evaluation run (have the agent produce proposed edits as patches) and only apply edits via a human-reviewed merge process.Like a lobster shell, security has layers — review code before you run it.
latestvk97f29czsg8tkbmt7g9zbqx6pn83gwcx
License
MIT-0
Free to use, modify, and redistribute. No attribution required.