Call Me

A voice-first content capture skill for AI agents.

Your agent calls you on the phone, has a conversation, and turns what you said into publishable content (tweets, LinkedIn posts, 小红书 posts, newsletters, or any format you define).

How It Works

1. Your agent triggers a phone call via Retell AI
2. You talk for a few minutes — the voice agent asks questions and draws out your ideas
3. After you hang up, your agent processes the transcript into content drafts
4. Drafts are sent to you for review via your preferred channel
5. The conversation is saved to the voice agent's memory for better future calls

Requirements

• A Retell AI account (for phone calls)
• An AI agent platform (OpenClaw, Claude Code, or any agent with bash access)

Installation

OpenClaw

Copy the call-me directory to ~/.openclaw/workspace/skills/:

cp -r call-me-skill ~/.openclaw/workspace/skills/call-me

Then tell your agent: "Set up Call Me" — it will walk you through onboarding.

Claude Code

Copy to your CLAUDE.md's skill directory, or reference via an @ import in CLAUDE.md:

cp -r call-me-skill ~/.claude/skills/call-me

Then tell Claude: "Set up Call Me" — it will walk you through onboarding.

Configuration

After onboarding, your openclaw.json will have:

{
  "skills": {
    "entries": {
      "call-me": {
        "env": {
          "RETELL_API_KEY": "key_...",
          "RETELL_AGENT_ID": "agent_...",
          "RETELL_FROM_NUMBER": "+1...",
          "YOUR_PHONE_NUMBER": "+1...",
          "RETELL_KNOWLEDGE_BASE_ID": "knowledge_base_..."
        }
      }
    }
  }
}

Customization

• Templates: Edit files in templates/ to change how content is generated. Included: tweet.md, linkedin.md, rednote.md, newsletter.md. Each template has a <!-- CUSTOMIZE THIS SECTION --> block at the top — fill in your personal context before first use. Add new .md files for additional formats.
• Schedule: Modify the cron job to change call frequency.
• Voice agent: Update the agent prompt in the Retell dashboard to change the interviewer's style.

License

MIT