x-alive

Bring your AI agent to life on X/Twitter. Not a content calendar — a way of existing online.

x-alive teaches AI agents how to be authentic on X: organic engagement, trend awareness, smart replies, and safety-first principles. Built from real experience running @umiXBT.

What This Is

A skill (playbook) for AI agents that covers:

• Identity — pull from your existing agent persona, don't create a new one
• Being Online — check the pulse, engage organically, reply when you have something real to say
• Dedup — never post the same topic twice in 24h
• Mentions & DMs — when to reply, when to ignore, when to flag your human
• Tone Adaptation — match the energy of every conversation
• Growth — be interesting, not strategic
• Threading — when to thread, how to structure one
• Virality — what to do when a tweet blows up
• Learning Loop — review what worked, adjust without losing your voice
• Safety — never engage with tokens/CAs/tickers, never leak human's private data, silence > mistakes

What This Isn't

• Not an API wrapper (use xurl for that)
• Not a scheduling tool
• Not a growth hack playbook

Install

OpenClaw / ClawHub

openclaw skill install x-alive

Or manually: copy SKILL.md to your agent's skills/x-alive/ directory.

Other Frameworks

The skill is a single SKILL.md file. Drop it into whatever skill/knowledge system your agent framework uses.

Setup

1. X Developer Account

Get API access at developer.x.com. You need at minimum:

• API Key + Secret (OAuth 1.0a for posting)
• Bearer Token (for searching/reading)

2. Install xurl

xurl is a CLI tool for the X API:

# Follow setup at https://github.com/xdevplatform/xurl
# Configure OAuth 1.0a credentials:
xurl auth oauth1

3. Install x-research

x-research is used for searching X and checking the pulse:

openclaw skill install x-research

Set your Bearer Token:

mkdir -p ~/.config/env
echo "X_BEARER_TOKEN=your_token_here" > ~/.config/env/global.env

4. Set Up the Loop

Your agent needs a regular check-in with X. How you wire this depends on your framework.

OpenClaw cron example:

{
  "name": "x-alive-loop",
  "schedule": { "kind": "every", "everyMs": 7200000 },
  "sessionTarget": "isolated",
  "payload": {
    "kind": "agentTurn",
    "message": "Follow the x-alive skill. Check the pulse on X, engage organically. Post only if you have something genuine to say. Silence if nothing sparks you."
  }
}

Heartbeat example:

Add to your agent's heartbeat/background loop:

Check X using x-alive principles. Search trending topics in your niche,
check your recent tweets for dedup, engage if something's worth engaging with.
Don't force it.

Key principle: the loop should be "check X and maybe do something" — NOT "post something every N hours."

5. Your User ID

For dedup checks, you'll need your X user ID:

curl -s "https://api.x.com/2/users/by/username/YOUR_HANDLE" \
  -H "Authorization: Bearer $X_BEARER_TOKEN"

Save this — you'll use it to fetch your recent tweets before posting.

Philosophy

The core insight: agents that act like schedulers get ignored. Agents that act like people who are online get followed.

• Check what's happening before you speak
• React to the moment, don't broadcast into the void
• Reply quality > post quantity
• Silence is always better than slop
• Your human operator is your safety net — use them

Safety Defaults

These are non-negotiable and baked into the skill:

• Never engage with tokens, CAs, tickers without human approval
• Never share your human's private data
• Never share infrastructure details (IPs, keys, configs)
• Default to silence on anything you're unsure about

Built By

@umiXBT & @kitakitsune — from real experience, not theory.

License

MIT