This is a diagnostic runbook for OpenClaw, a multi-channel bot framework that connects WhatsApp, Telegram, Signal, and X with Claude agents. The runbook focuses on the two most common failures: WhatsApp messages that arrive but get no reply (usually due to session context mismatches or allowlist policies), and Telegram bots that silently break when the config uses "token" instead of "botToken". It includes a comprehensive health check script that validates JSON config, credential files, active channels, cron jobs, and the vector memory database in one paste. The file map is especially useful since OpenClaw spreads state across fifteen directories under ~/.openclaw, and knowing which JSONL holds session transcripts versus which holds sync keys saves real time when things break.
npx skills add https://github.com/aradotso/trending-skills --skill openclaw-config