How to use the Twitter/X Management Expert agent skill

What the skill makes the agent do

A skill is a folder of instructions that an AI agent reads when a task matches it. This one is prompt-only: it contains knowledge and examples, not programs. Its SKILL.md file gives the agent a Twitter API v2 reference with ready-made curl commands, a content strategy framework, tweet writing rules, hashtag guidance, and rules for replying to mentions.

Once installed, the agent can draft posts, plan a week of content and work out how to carry out account tasks. Acting on the account still depends on the agent having access to a Twitter API token.

• Content pillars: 3-5 core topics that all posts revolve around
• A weekly mix of seven post types, such as hot takes, threads, tips and questions
• Suggested posting times by day, adjustable to your audience's timezone
• Hook and thread writing rules, and a 280-character limit reminder
• Hashtag guidance: 0-2 per tweet, and none in threads except possibly tweet 1
• A playbook for replying to mentions, including reply templates and when not to engage

Install the skill from the ZIP download

Download the ZIP for this skill from this page and unzip it. The skill folder contains two files: HAND.toml and SKILL.md. Place the unzipped folder in one of these locations, depending on how you want to use it:

• Personal use across projects: ~/.claude/skills/
• A single project: the .claude/skills/ folder inside that project
• Claude apps: upload the ZIP under Settings → Capabilities → Skills

Set up the credentials the skill needs

The skill calls the Twitter API, so the agent needs a Twitter/X developer account with API access. The file expects your bearer token to be available as the environment variable TWITTER_BEARER_TOKEN. The agent also needs curl or another HTTP client to make requests.

Actions performed as your user, such as posting, deleting and liking, need User-context OAuth 1.0a credentials. The skill mentions these but does not set them up for you, so you will need to configure them separately.

• Twitter/X developer account and API access
• TWITTER_BEARER_TOKEN environment variable
• curl or another HTTP client the agent can run
• User-context OAuth 1.0a credentials for user actions (not configured by the skill)

Trigger the skill with natural requests

You do not need special syntax. Describe the task in plain language that matches what the skill covers, such as planning posts, drafting a thread, or answering mentions. The agent will use the skill's reference and rules when the request fits.

Ask for a draft first. The skill has no approval step, so review any text before it is published.

• Draft a week of posts for my pillars: AI tools, startup building and engineering culture.
• Write a thread about what I learned launching my product, using the thread structure from the skill.
• Check my last 10 mentions and draft replies I can review before posting.
• Look up the metrics for these three tweet IDs and tell me which performed best.
• Suggest a posting schedule for next week based on the suggested times.

Safety notes: scripts and network access

The skill runs no scripts. It is made of instructions and curl examples, and it does use the network: every API call goes to api.twitter.com using your token. Posts, replies, likes and deletions are public actions on your account and are hard to reverse.

The skill does not cover Twitter/X platform policy or spam limits. Automated engagement can break platform rules or harm your reputation, so check the rules yourself before letting an agent like or reply at volume.

• Review every draft before it is posted, liked or deleted
• Use a token with only the access you need
• Check the endpoints and rate limits against current Twitter/X documentation, since the file may be out of date
• Never auto-like controversial or political content, per the skill's own rules

Tips for getting useful results

The skill is most useful when you give it context about your audience and goals. The reference is general, so the more specific your pillars, voice and examples, the better the drafts. Use the rate-limit headers the skill describes to avoid hitting limits during batch work.

• Name your 3-5 pillars and rough percentages before asking for a calendar
• Give a sample of your own posts so the agent can match your voice
• Ask for reply templates adapted to each mention rather than copying one template
• Track results with the metrics endpoint: like_count, reply_count, impression_count and similar fields