openclaw-customizer_skill

Overview

This skill guides you through configuring, customizing, and creatively leveraging OpenClaw — the self-hosted AI gateway that connects LLMs to messaging channels. It helps edit openclaw.json, author bootstrap files (SOUL.md, USER.md, AGENTS.md, IDENTITY.md, TOOLS.md), set up channels, build multi-agent routing, and create skills/hooks/cron jobs with concrete examples.

How this skill works

I inspect the relevant workspace and config files (openclaw.json, workspace bootstrap files, skills, hooks, cron jobs) and propose concrete JSON5 or markdown edits you can paste. For troubleshooting I recommend running openclaw doctor and point to logs (/tmp/openclaw/openclaw-YYYY-MM-DD.log). I always include why each change is needed and adjacent improvements like failover providers or memory pipelines.

When to use it

• Setting up or modifying ~/.openclaw/openclaw.json or per-agent configs
• Writing or improving SOUL.md, USER.md, AGENTS.md, IDENTITY.md, TOOLS.md bootstraps
• Connecting or debugging messaging channels (Telegram, WhatsApp, Discord, Slack, iMessage)
• Designing multi-agent routing, bindings, and workspace isolation
• Creating skills, hooks, cron jobs, or automated pipelines
• Troubleshooting errors, authentication, or model/provider failover

Best practices

• Always provide a JSON5 snippet for openclaw.json edits; include comments explaining keys
• Keep agent workspaces isolated and separate credentials per workspace to avoid cross-agent leaks
• Use allowFrom policies — never set an open DM policy without explicit warning and reasons
• Include timezone in cron jobs and show sample job entry (e.g., timezone: 'UTC')
• Add a lightweight heartbeat hook for routing health checks and provider failover

Example use cases

• Multi-channel personal assistant: route Telegram for notifications, Slack for team tasks, and iMessage for personal DMs using per-channel personas
• Cron-driven daily digest: cron job reads workspace MEMORY.md, summarizes events, and posts to Discord at 09:00 America/New_York
• Provider failover: primary LLM provider with a secondary fallback and automatic retries in openclaw.json
• Skill + hook automation: skill extracts calendar events, hook triggers reminder messages via WhatsApp
• Creative pipeline: agent A produces research notes, agent B refines them into publish-ready summaries and posts to Slack