How to use the Channel Welcome Onboarding agent skill (welcome)

What the welcome skill makes the agent do

An agent skill is a folder of instructions that an agent reads to decide how to handle a particular task. This one covers the first conversation on a newly wired channel. The skill is triggered automatically when a channel is first wired, and it tells the agent to send a friendly greeting, state its name, and offer to show what it can do.

Rather than listing every capability at once, the skill asks the agent to reveal them one at a time, in a set order: memory, spawning helper agents, scheduled tasks, web research, coding, interactive buttons, file output, and adding tools. It then explains that sensitive actions need the user's approval and that nobody can talk to the agent without the owner's permission.

• It does not add new abilities to the agent. It only shapes the first conversation.
• The agent improvises wording, so the message will vary between runs.
• It also asks the agent to note the user's recurring domains and feedback in its memory files.

Install the welcome skill

Download the ZIP from the skill page on this site. Then choose one of two options depending on where you want the skill to apply.

For a personal setup, unzip the file into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.

The package contains one file, SKILL.md, so after unzipping you should see the welcome folder with that file inside it.

• Personal use: ~/.claude/skills/
• Single project: the project's .claude/skills/ folder
• Claude apps: Settings → Capabilities → Skills, then upload the skill

How the skill gets triggered

The skill is designed to run automatically when a channel is first wired, so you normally do not type a command to start it. The source says there are no special commands: users just talk naturally, and if they want something done, they say so.

If the instruction that triggered the welcome names a channel addendum file, such as one for Slack, the agent reads that file first and follows it. The addendum may replace a section of the welcome or add steps. If no addendum is named, the agent runs the skill exactly as written.

• Automatic: the trigger fires when a channel is first wired.
• Addendum: an optional channel-specific file, for example addenda/slack.md, adjusts the welcome for that channel.
• Without an addendum: the standard welcome runs unchanged.

Example flow of a first conversation

After the greeting, the agent asks whether the user wants to explore its abilities or jump straight into something. If they choose to explore, it introduces one capability at a time, briefly explains it, and offers a demo or a chance to try it. If they choose to jump in, it simply starts the task.

Sensitive actions such as installing packages or adding MCP servers need the user's explicit approval first. The user also controls who can talk to the agent: adding it to a new group or sharing a bot link triggers an approval request on the user's side.

Confirmations and corrections the user gives during onboarding are saved to memory for later sessions.

• Ask for the user's preferences, not a full manual.
• Expect approval prompts before package installs or new MCP servers.
• Corrections given during onboarding are stored as feedback.

What the skill needs and what to watch for

The skill lists no requirements. It does not run scripts and does not use the network, according to its listing. Because it is only instruction text, its behaviour depends on the agent that reads it and on the capabilities that agent actually has.

Do not use it where the onboarding text must match a fixed legal or compliance script. The agent improvises wording and may overstate capabilities it does not have. Review the first messages on a new channel before relying on them.

• Requirements: none listed.
• Scripts: none run.
• Network: none used by the skill itself.
• Check the wording and the capability claims on your own channels before going live.

Tips for a better first run

The skill asks the agent to scan its available MCP tools and skills before starting, so the capabilities it describes match its real setup. If you add or remove tools, review the welcome again.

Keep the channel's tone in mind. The skill asks for casual wording on consumer chat apps and somewhat more professional wording on workplace platforms. A channel addendum is the place to adjust this for a specific channel.

• Match tools to claims: keep the agent's real capabilities in line with what it introduces.
• Use a channel addendum for channel-specific tone or steps.
• Read the memory files the agent writes to confirm they hold what you expect.