AI Agent Skills: Install Gmail, GitHub, Stripe, and 50+ Integrations in One Command

2026-03-02 · 6 min read · Developer · 0 views

What a "skill" is (and why it matters)

A skill is a self-contained module that gives your AI agent the ability to use a specific tool or service. Under the hood, it's a SKILL.md file (instructions for the agent on when and how to use the tool) plus any scripts needed to actually execute actions.

The SKILL.md is the key insight. Rather than writing complex system prompts or hardcoding tool behavior, you describe the tool's capabilities, constraints, and best practices in plain language. The agent reads this before acting, which means it uses the tool correctly, handles edge cases, and knows when NOT to use it.

Installing from the ClawHub marketplace

The ClawHub marketplace has 50+ production-ready skills. Installation is one command: clawhub install gmail. That's it. The skill is added to your agent's toolset immediately -- no restart, no configuration.

The most-installed skills: Gmail (read, send, draft, search), Google Calendar (create events, check availability, reschedule), GitHub (create PRs, review code, manage issues), Stripe (query MRR, check subscriptions, process refunds), Slack (post messages, search history, manage channels), Notion (create pages, update databases, search), Airtable (query and update records), Twilio (send SMS), Apple Notes/Reminders (native Mac integration), Things 3 (full task management).

BYOK: your keys, your data, your control

Every skill uses your own API credentials -- not ours. When the Gmail skill sends an email, it's authenticating with your Google OAuth token stored on your private server. When the Stripe skill queries MRR, it's hitting the Stripe API with your secret key.

This means your data never passes through any intermediary. OpenClawInstaller.ai never sees your emails, your code, or your financial data. The agent runs on your server, calls the APIs with your keys, and the results stay in your environment.

Skill chaining: the real power

Skills compound when chained. "Summarize my unread emails, add action items to Things 3, and post a Slack summary to #standup" touches three skills in one request. The agent reads the SKILL.md for each, executes them in sequence, and handles the data handoff between steps.

More complex chains are possible: monitor GitHub for new issues -> research similar issues -> draft a response -> post to the issue -> notify Slack. This runs as a scheduled cron job. You never touch it after setup.

Writing a custom skill in 15 minutes

Any API becomes a skill. You need three things: a SKILL.md describing the tool and when to use it, an API script (usually under 50 lines of bash or Python), and a test confirming it works. The agent reads the SKILL.md, decides when the skill applies, and calls the script when needed.

Skills can be published back to ClawHub for others to use. The ecosystem grows with every developer who builds their own integration.

Get your agent running with every tool you use ->