Twitter Monitor

Workflow

Use this skill as an execution workflow, not as a long-running daemon. First complete a one-shot fetch, then offer optional Feishu sync and scheduling only when useful.

1. Ask for X/Twitter account ids when missing. Accept handles such as sama, @sama, profile URLs, or multiple comma-separated ids.
2. Ask the user to provide or configure a twitterapi.io API key when TWITTER_API_KEY is unavailable. Do not write API keys into files committed to a repository.
3. Confirm pagination depth. Default to --pages 1; use a larger number only when the user asks for more history or accepts higher API usage.
4. Run scripts/twitter_monitor.py and generate JSON or CSV output.
5. If the user wants Feishu/Lark Bitable output, follow references/feishu-output.md.
6. After the one-shot command works, ask whether they want recurring execution. If yes, follow references/scheduling.md, including OpenClaw when appropriate.

Quick Start

From this skill directory:

export TWITTER_API_KEY="..."
python scripts/twitter_monitor.py --accounts sama --pages 1 --format json

For multiple accounts:

python scripts/twitter_monitor.py --accounts sama,elonmusk,OpenAI --pages 1 --format csv

For a JSON account list:

python scripts/twitter_monitor.py --accounts-file assets/accounts.example.json --pages 1

Read references/twitterapi-setup.md when the user needs API key setup, account input examples, or command variants.

Output Schema

The script writes one record per tweet with these fields:

• 推文内容
• 日期
• 推文链接
• 推文ID
• 作者
• 作者ID
• 阅读量
• 点赞数
• 转发数
• 评论数
• 收藏数
• 是否回复
• 抓取时间

Deduplicate by 推文ID before appending to any durable destination.

Feishu

Do not require Feishu configuration for normal fetches. Only enter the Feishu workflow when the user asks to write, append, or sync records to Feishu/Lark Bitable.

Prefer feishu-cli over hard-coded Feishu app credentials. If feishu-cli is missing, ask whether to install and configure it. See references/feishu-output.md for the field mapping and sync rules.

Scheduling

Offer scheduling only after a successful one-shot fetch or when the user explicitly asks for ongoing monitoring.

Prefer OpenClaw when the user wants an agent-managed recurring job or already uses OpenClaw. Otherwise choose Codex automations, cron, or launchd based on the runtime environment. See references/scheduling.md.

Safety

Never commit real API keys, Feishu tokens, output files containing private monitoring data, or local status caches. Keep secrets in environment variables, secret managers, or the user's existing CLI auth.