自我复盘

自我复盘与持续改进技能。当用户要求"复盘"、"总结经验"、"记录教训"、 "自我提升"、"持续改进"、"错题本"、"学习日志"时触发。 主动在每次完成任务、犯错、学到新知后，将内容写入 reflections/。

What to consider before installing/running: - This skill runs entirely on your machine and does not ask for API keys or network access, but it will read/write files under ~/.openclaw/workspace and may read /tmp/openclaw logs. If those directories contain sensitive data, review what will be written. - The daily_reflect script can create a launchd plist and call launchctl load to schedule itself daily. Only run the --setup flow if you want a persistent user-level scheduled task; inspect the generated plist before loading. - auto_remind reads a hard-coded /tmp/openclaw/openclaw-<date>.log path to detect command failures. Ensure that path is appropriate for your environment (or empty) to avoid unintended log parsing. - The code contains no network calls today; a commented note mentions Feishu integration but that is not implemented. If you pull updates to this skill in future, re-check for added network/client code. - Recommended steps: (1) review the scripts locally, (2) run python3 scripts/auto_remind.py --check and python3 scripts/repeat_detect.py manually to see behavior, (3) if you want scheduled runs, use python3 scripts/daily_reflect.py --setup and confirm the plist content before loading, (4) back up any existing reflections/memory files before enabling automated writes.

Type: OpenClaw Skill Name: self-reflection-lan Version: 1.0.1 The skill bundle implements a self-reflection and logging system that includes high-risk behaviors, specifically the creation of a persistence mechanism via macOS launchd (LaunchAgents) in scripts/daily_reflect.py. It also uses os.system for shell command execution, which is a significant vulnerability. While these actions are aligned with the stated purpose of scheduling daily tasks, the combination of startup persistence and unsanitized shell execution poses a security risk.