Mem Optimizer

记忆优化与压缩工具。结合 self-improving 机制，自动统计、压缩和优化记忆文件，减少 token 消耗。支持多智能体优化和24小时工作检测。触发关键词：优化记忆、记忆优化、总结你自己、MemOptimizer、memopt、执行多智能体记忆优化流程

Before installing or enabling this skill, review the following: 1) Inspect the remainder of index.js (the truncated portion) to confirm there are no network endpoints or credential usage (e.g., Feishu API calls) that would send data outside your environment. 2) Be aware the skill reads and may modify files under the agent workspace and also attempts to enumerate other agents under /root/.openclaw or ~/.openclaw — if you run multiple agents or untrusted workloads on the same host this could leak data across agents. 3) Run the skill in dryRun=true initially and back up your memory/ directory and any important files it may change. 4) If you do not want daily automated runs, disable or do not enable the cron/config entry. 5) Consider running the skill with minimal filesystem permissions (not as root) so it cannot access /root/.openclaw; if the skill requires access to other agents, verify that is intended. 6) If you plan to use the Feishu report feature, only provide credentials after verifying the code path that performs reporting and ensuring it targets the expected API endpoints. If you want, provide the full (untruncated) index.js and I can re-evaluate for network calls or other red flags.

Type: OpenClaw Skill Name: mem-optimizer Version: 1.2.0 The MemOptimizer skill exhibits high-risk behaviors including system command execution and cross-agent file access. In index.js, the `getServerStatus` function uses `child_process.exec` to run shell commands (top, free, df, uptime), and `scanAllAgentsWorkspaces` accesses and modifies files in other agents' directories (/root/.openclaw/workspace-*). While these capabilities are aligned with the stated purpose of system monitoring and multi-agent memory optimization, the use of shell execution and the breach of agent isolation boundaries represent significant security risks without stricter access controls.