龙虾记忆大师

This skill appears to implement what it claims (local long‑term memory files and simple archive/check scripts), but take these precautions before installing or enabling autonomous behavior: - Confirm the workspace location: the scripts use OPENCLAW_WORKSPACE and default to /root/.openclaw/workspace. Set OPENCLAW_WORKSPACE to a directory you control (not /root) and verify permissions. - Review and approve any instruction that creates system cron jobs or AGENTS.md; scheduling tasks or creating agent policies grants ongoing behavior and may run outside your immediate control. - Understand that the skill will write and modify files under the workspace (MEMORY.md, memory/*.md, SESSION-STATE.md, working-buffer.md). If these files contain private data, consider encrypting or limiting access. - The included scripts are short and do not perform network calls or read secrets, but you should inspect them (they are provided) and run them in a sandbox first. - Recommend asking the publisher to declare OPENCLAW_WORKSPACE in requires.env or document the expected workspace explicitly, and to clarify how AGENTS.md and cron entries should be managed and revoked. If you need the skill but want to minimize risk: run it in a non-privileged, isolated workspace with explicit environment variables, and do not enable system cron/agent automation until you’ve inspected AGENTS.md and confirmed intended behaviors.