About application-skills

application-skills is a library of 3,000+ integration skills designed to empower AI agents to connect with third-party applications like Gmail, Slack, HubSpot, Salesforce, GitHub, and thousands more. Built on the Membrane platform and compatible with the open Agent Skills specification, these skills teach agents how to interact with specific apps by defining their structure, available actions, and usage context. The package provides drop-in support for agents including OpenClaw, Cursor, Claude Code, GitHub Copilot, and Gemini CLI, allowing them to perform complex tasks such as sending messages, creating tickets, or updating spreadsheets without requiring API keys in prompts or manual documentation lookups. Membrane automatically manages authentication credentials including OAuth and token refreshes, ensuring secure and seamless connections. Each skill is organized as a dedicated module containing detailed instructions on app concepts and actions. Users can easily install specific capabilities by referencing t

m

Published by

membranedev

Visit View Profile

README.md

View on GitHub

3,000+ Integration Skills by Membrane

Give your AI agent skills to connect with Gmail, Slack, HubSpot, Salesforce, GitHub, and 3,000+ other apps.

Each skill teaches your agent how to work with a specific app: its structure, actions, and when to use what. Built on Membrane and the open Agent Skills specification.

What’s in this repo?

3,000+ app-specific skills — One skill per app (e.g. gmail, slack, hubspot, salesforce). Each skill describes the app’s concepts, available actions, and how to use the Membrane CLI.
Works with your agent — Drop-in support for OpenClaw, Cursor, Claude Code, GitHub Copilot, Gemini CLI, and any agent that supports the Agent Skills spec. Use with OpenClaw’s ClawHub or install via npx skills add.
Auth handled for you — OAuth, API keys, token refresh: Membrane manages credentials so your agent doesn’t have to.

What your agent can do

Once a skill for an App is installed, your agent can connect to that app and run actions on your behalf — no API keys in the prompt, no copy-pasting docs.

You say	Agent does
"Send a message to #general on Slack: deploy is done"	Uses the Slack skill, finds the right action, runs it with the right params.
"Create a Jira ticket for the login bug, assign to me, high priority"	Uses the Jira skill, creates the issue, sets assignee and priority.
"Add a new row to our Google Sheet with today’s metrics"	Uses the Google Sheets skill, finds the connection, runs the action.
"List my last 10 emails from Gmail"	Uses the Gmail skill, lists messages.

Each skill includes the app’s structure (entities and resources), available actions, and when to use which action when it’s not obvious.

Skill layout

Skills are organized as:

skills/
  gmail/       → SKILL.md
  slack/      → SKILL.md
  hubspot/    → SKILL.md
  salesforce/ → SKILL.md
  ...         (3,000+ more)

Install by skill name = folder name (e.g. gmail, slack, hubspot).

Popular skills (examples)

App	Skill name
Gmail	`gmail`
Slack	`slack`
HubSpot	`hubspot`
Salesforce	`salesforce`
GitHub	`github`
Jira	`jira`
Google Drive	`google-drive`
Dropbox	`dropbox`
Pipedrive	`pipedrive`

Browse the skills/ folder for the full list.

Resources

Membrane — Integrations API for AI Agents

application-skills