Xiaopi Self Improving

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。

This skill appears to be a simple local recorder for errors, corrections, and best practices — the shipped scripts match that. However, SKILL.md claims automatic cross-project syncing and automatic captures that are not implemented in the code; do not assume the skill will safely and automatically sanitize or sync memories. Before installing or enabling autonomous invocation: 1) Inspect and test the scripts in a safe environment; run them manually to see what gets written under ~/.openclaw/memory/self-improving. 2) Avoid passing secrets in commands or messages the skill might record; consider adding or requesting sanitization before use. 3) If you expect 'automatic' behavior (capture on every failed command or every chat correction), realize additional integration code or agent policies would be required — otherwise nothing will be auto-triggered. 4) If you want automatic syncing to project files or AGENTS.md, ask the author for a clear, auditable implementation (or implement it yourself) rather than relying on the SKILL.md claims. 5) If enabling autonomous agent invocation, restrict when the agent can call these scripts and review recorded files regularly (or put the memory directory under a repo with review) to reduce accidental data leakage.

Type: OpenClaw Skill Name: xiaopi-self-improving Version: 1.0.0 The skill implements a local memory and self-improvement system for an AI agent by logging command errors, user corrections, and best practices to JSONL files in the user's home directory (~/.openclaw/memory/self-improving). The provided Python scripts (log_error.py, log_correction.py, log_best_practice.py, and check_memory.py) use standard libraries to perform basic file I/O and do not exhibit any signs of data exfiltration, unauthorized execution, or malicious intent. The instructions in SKILL.md are consistent with the stated goal of helping the agent learn from past interactions.