How to update OpenClaw and manage skills from Telegram

Once your OpenClaw agent is running, the Telegram bot is your primary interface — no SSH required for day-to-day operations. This post covers how to keep your agent current and extend it with new skills entirely from Telegram.

When ClawKickstart provisions OpenClaw, it sets the Telegram channel to open DM mode so your messages go straight to the agent. OpenClaw routes your instructions through its Terminal Backend, which means shell commands run on the remote server at your direction — without you ever opening a terminal session.

• /help — Shows all available commands and a brief description of each
• /status — Reports the agent's current provider, model in use, and health state
• /tasks — Lists background tasks running in the current session
• /tools — Shows all tools the agent currently has loaded and available
• /stop — Halts any running task immediately
• /approve <id> <decision> — Resolves any pending execution-approval prompts

Before updating, it is good practice to ask your agent to make a backup of your configuration file first. Tell it: "Make a backup copy of my openclaw.json file so I can revert if something breaks." The agent will copy ~/.openclaw/openclaw.json to ~/.openclaw/openclaw.json.bak. Once you have a backup, tell the agent: "Run openclaw update to update yourself to the latest version." It will pull the latest changes, run openclaw doctor to migrate any configuration schema changes, and restart the gateway. You will receive a confirmation in the chat when done.

# 1. Ask the agent to back up your config first
cp ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.bak

# 2. Then update
openclaw update
openclaw doctor
openclaw gateway restart

OpenClaw skills extend what the agent can do. The official community registry is ClawHub (clawhub.ai) — a growing library of skills for GitHub workflows, research pipelines, productivity automations, and more. You can browse and install skills entirely from Telegram by asking the agent to search and install them on your behalf.

Tell your agent: "Search ClawHub for skills related to GitHub workflows and install the best match." The agent will run the search, show you options in the chat, and install your chosen skill. A word of caution: anyone can publish to ClawHub, so some skills may be low-quality or unsafe. Always ask your agent to run openclaw skills info <slug> before installing anything — it will pull the SKILL.md and show you exactly what the skill does and what permissions it requires.

openclaw skills list              # See what is installed
openclaw skills search "<query>"  # Browse ClawHub registry
openclaw skills info <slug>       # Always inspect before installing
openclaw skills install <slug>    # Install a verified skill by slug
openclaw skills update --all      # Update all installed skills
openclaw gateway restart          # Apply changes after install

There is a shortcut for power users: if you find a skill repository or SKILL.md file online that you want to use, you can drop the URL into Telegram and ask your agent to recreate it. For example: "Here is this skill from GitHub — please read it and recreate it in my skills folder." This works particularly well with OpenAI and Anthropic models, which are strong at reading and rewriting structured documents. The agent will create a local copy in ~/.openclaw/skills/ that you own and can freely modify.

• Workspace skills (<workspace>/skills/) — highest priority, project-specific overrides
• User skills (~/.openclaw/skills/) — user-wide, installed via CLI or recreated from URLs
• Bundled skills — shipped with OpenClaw itself (lowest priority)

Using Telegram to manage your OpenClaw agent means you never have to leave your phone or your chat client to keep your agent healthy and up to date. Updates, skill installs, backups, and health checks all flow through the same interface you already use to talk to the agent.

ClawKickstart provisions OpenClaw on your own VPS with Telegram already configured and the gateway running, so you can use all of these commands from day one — no manual gateway setup required.