AI 模型智能管家

What to consider before installing: - Metadata mismatch: the registry summary showed no required env vars, but skill.json and the code require OPENROUTER_API_KEY (and optionally Feishu tokens). Don't trust the top-line registry claim — verify the skill.json and README. - Back up your OpenClaw config (~/.openclaw/openclaw.json) before installing or running this tool. The deployer will read, backup, and overwrite that file to add/remove fallback models. - Review the code yourself (or have someone you trust audit it). The package is pure Python and readable; check any missing modules (e.g., bitable_writer referenced but not included) and confirm no hidden network endpoints. - Limit credentials: if you provide OPENROUTER_API_KEY or Feishu app tokens, prefer creating least-privilege tokens or using ephemeral/test accounts first. The tool will read tokens from your openclaw.json or prompt for them. - Run in an isolated environment first (VM or non-production user account) to confirm behavior (what files are written, how backups are made, what network calls occur). - If you plan to allow automated cron runs, be aware that scheduled execution plus the ability to modify global config increases blast radius—ensure you trust the author or have internal approval processes in place. If you want, I can: (a) list the exact lines that read/write ~/.openclaw/openclaw.json, (b) show where the code prompts for or uses credentials, or (c) suggest a minimal safe test plan to run the tool in isolation.