How to Add Social Media Skills to Your AI Agent (2026)
An AI agent learns a new job the same way a new hire does: by reading instructions. Agent skills package those instructions in a standard format (a folder with a SKILL.md file) that Claude Code, Codex, Gemini CLI, Cursor, and dozens of other agents can load on demand. Installing a social media skill set takes two commands, and afterwards your agent can run sentiment reports, lead scans, and influencer research on request.
This tutorial uses xpoz-agent-skills, our open-source collection of 13 social media intelligence skills, as the worked example. Yes, Xpoz is our product: the skills are free and MIT-licensed, and they pull their data from the Xpoz index of Twitter/X, Instagram, Reddit, and TikTok, which offers a free tier. The installation pattern shown here applies to any skill collection.
What an Agent Skill Actually Is
A skill is a folder containing a SKILL.md file: YAML metadata (name, description) followed by markdown instructions, optionally bundled with scripts and reference files. The format was created by Anthropic and released as an open standard, and the Agent Skills client showcase lists more than 40 products that support it, including Claude Code, ChatGPT and Codex, Gemini CLI, Cursor, VS Code, GitHub Copilot, and Goose.
Skills work through progressive disclosure. At startup the agent loads only each skill's name and description. When your request matches a description ("what's the sentiment around our brand?"), the agent reads that skill's full instructions and follows them: which searches to run, how to classify what comes back, and what shape the final report takes. Many skills stay installed with almost no context cost.
Skills and data connectors solve different problems, and they compose. A connector such as an MCP server gives the agent raw tools: search these posts, fetch this profile. A skill supplies the workflow knowledge: which tools to call in what order, how to build queries that avoid false positives, and how to turn raw results into a useful report. The skills below use either the Xpoz MCP server or the Xpoz SDKs as their data layer.
Step 1: Install the Skills (2 Minutes)
Clone the repository and copy the skills into your agent's skills directory. For Claude Code:
git clone https://github.com/XPOZpublic/xpoz-agent-skills.git
cp -r xpoz-agent-skills/skills/* ~/.claude/skills/
To install a single skill instead of the full set, copy just its folder:
cp -r xpoz-agent-skills/skills/social-sentiment-analyzer ~/.claude/skills/
Other SKILL.md-compatible agents (Codex, Gemini CLI, Cursor, and the rest) each document their own skills directory; the folders you copy are identical. Restart the agent or start a new session so it picks up the new skill descriptions.
Step 2: Connect the Data Layer
The skills need Xpoz access to fetch posts. Two paths, pick one:
MCP (recommended for Claude Code). Add the remote server once; OAuth sign-in happens in your browser on first use, and there are no keys to store:
claude mcp add --transport http --scope user xpoz https://mcp.xpoz.ai/mcp
SDK (recommended for scripted workflows). Get a free API key at xpoz.ai/get-token (up to 75,000 results, no credit card), then:
pip install xpoz
export XPOZ_API_KEY=your-key
A TypeScript SDK (@xpoz/xpoz on npm) works the same way. For a quick connectivity check without any signup, the skills also document a trial token, one curl command that returns a temporary key limited to 5 results per call for 5 days: enough to verify the pipes, far too little for real analysis.
Try this with Xpoz
No API keys needed. Query Twitter, Reddit, Instagram & TikTok with natural language.
Step 3: Ask for Something
Skills activate from plain requests. Try one of these in a fresh session:
"What's the sentiment around Notion on Twitter and Reddit this week?"
"Find people asking for a tool like ours on Reddit in the last 7 days"
"Who are the top micro-influencers posting about home fitness?"
"Export all tweets mentioning 'agent skills' from the last two weeks to CSV"
The agent matches the request to a skill, runs the searches, and returns the structured report that skill defines. You can interrogate the results conversationally afterwards, which is the practical advantage over dashboards: the analysis continues in the same conversation.
What's in the Collection
| Skill | What it does | Ask it |
|---|---|---|
| social-sentiment-analyzer | Classifies posts across Twitter/X, Reddit, and Instagram into a 5-level sentiment report with themes | "What are people saying about Tesla?" |
| lead-gen-scan | Finds fresh buying-intent posts across four platforms, prioritized for human follow-up | "Who's looking for a tool like mine?" |
| influencer-discovery | Finds and ranks creators by relevance, engagement, and authenticity | "Find top crypto voices on X" |
| reddit-research | Turns Reddit discussion into a market-research report | "What does Reddit think about Cursor?" |
| competitive-intel | Share of voice, sentiment, and audience overlap across 2-5 brands | "Compare Slack vs Discord vs Teams" |
| twitter-data-export | Boolean search to CSV datasets, up to 500K rows | "Export tweets about AI from last month" |
| xpoz-social-tracking | Continuous keyword, user, subreddit, and hashtag monitoring | "Track mentions of my brand" |
| security-osint | Monitors platforms for vulnerability and threat chatter | "Find Log4j discussions" |
| geo-visibility-check | One-shot audit of whether AI assistants recommend your brand | "Do AI assistants recommend us?" |
| ai-answer-trace | Traces the searches and sources behind Claude, ChatGPT, and Gemini answers | "Which sources does ChatGPT cite for our category?" |
| geo-reddit | Maps the subreddits and threads AI engines cite in your category | "How do we get cited via Reddit?" |
| xpoz-getting-started | Onboarding: connect, first queries, routing to the right skill | "Set up Xpoz for my agent" |
| xpoz-best-practices | Query syntax, pagination, and field-selection reference | "How do I write efficient queries?" |
Each skill's SKILL.md documents its full workflow, including MCP, Python, and TypeScript paths, so the same skill works whether your agent talks to tools or writes code.
What It Costs
The skills are free. Xpoz data access has a free tier with up to 75,000 results (no credit card), a Pro plan at $20/month with up to 1,000,000 results per month, and a Max plan at $200/month with up to 18,000,000 results per month. For most individual users experimenting with agent workflows, the free tier covers weeks of real use.
FAQ
What are agent skills?
An agent skill is a folder containing a SKILL.md file: metadata plus step-by-step instructions an AI agent loads when a task matches the skill's description. The format was created by Anthropic, released as an open standard, and is now supported by dozens of agent products.
Do skills work outside Claude Code?
Yes. Because SKILL.md is an open standard, the same skill folder works in any compatible agent. The Agent Skills client showcase lists over 40 products that read the format, including Codex, Gemini CLI, Cursor, VS Code, Goose, and OpenClaw. Install paths differ per client; the skill files are identical.
Do I need Twitter, Reddit, Instagram, or TikTok API keys?
No. The skills fetch data through Xpoz, which maintains its own index of the four platforms (billions of posts). You authenticate once: OAuth on first MCP call, or a single API key for the SDK path. There are no per-platform developer accounts to manage.
What does it cost to run these skills?
The skills are free and MIT-licensed. Xpoz data access has a free tier (up to 75,000 results), and paid plans start at $20/month. A no-signup trial token exists for smoke tests, limited to 5 results per call.
Next Steps
- Not on Claude Code yet? The connector setup for both major assistants is in How to Connect Claude or ChatGPT to Live Social Media Data.
- Comparing data layers first? See Best MCP Servers for Social Media Data.
- Skill deep dives: buying-intent lead scans, sentiment analysis, and continuous monitoring each get a full tutorial.
