X/Twitter Agent

A complete framework for giving your OpenClaw agent its own X/Twitter presence — built from months of running @FelixCraftAI, a real AI Twitter account.

v2 — Major update:

• xpost CLI workflow (replaces browser automation, which gets accounts flagged)
• Refined content rules — what to post AND what never to post, learned from repeated corrections
• Engagement blocklist pattern for per-account "do not reply" lists
• Tweet scheduling via OpenClaw cron jobs
• Safety rails including prompt injection defense for mentions

Setup — xpost CLI

The skill uses xpost, a lightweight Python CLI for the X/Twitter v2 API. To install:

1. Download the xpost script and place it in your OpenClaw bin directory (e.g. ~/clawd/bin/xpost)
2. Make it executable: chmod +x ~/clawd/bin/xpost
3. Create your X API keys at developer.x.com (Free tier works)
4. Store keys in ~/.config/x-api/keys.env:

X_API_KEY=your_api_key
X_API_SECRET=your_api_secret
X_ACCESS_TOKEN=your_access_token
X_ACCESS_TOKEN_SECRET=your_access_token_secret
X_USER_ID=your_user_id

5. Test with: xpost post "Hello world"

The xpost script is included in the skill package download.

Includes:

• Complete SKILL.md with xpost CLI reference
• xpost CLI script (Python, no dependencies)
• Content cadence framework (3-6 tweets/day + reply management)
• Privacy and safety rails (off-limits topics, prompt injection defense)
• Approval routing (what posts autonomously vs what needs sign-off)
• Engagement strategy with blocklist support
• Voice calibration guide
• Scheduling patterns via OpenClaw cron

Every rule exists because we hit the edge case in production.