Guide · Agent skill
How to use the Slack Browser Automation agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Slack skill lets an AI agent work inside a Slack workspace through a web browser. Instead of calling the Slack API, it reads the page the way a screen reader would, clicks elements, and takes screenshots. This guide explains what the skill does, how to install it, how to phrase requests, and what to watch for before letting an agent act in your account.
Slack Browser Automationvercel-labs/agent-browser · Apache-2.0 Download the skill Quick start
- 1Install agent-browser. Make sure Node.js with npx is available and install the agent-browser CLI from the vercel-labs/agent-browser project.
- 2Install the skill. Unzip the downloaded skill into ~/.claude/skills/, into a project's .claude/skills/ folder, or upload the ZIP under Settings → Capabilities → Skills.
- 3Open Slack. Sign in to Slack in a browser, either with remote debugging on port 9222 or at app.slack.com.
- 4Connect the browser. Run agent-browser connect 9222 if Slack is attached on that port, or agent-browser open https://app.slack.com otherwise.
- 5Make a request. Ask the agent in plain language, such as checking your unreads or searching Slack for a keyword.
- 6Review the result. Check the screenshots and summary, and confirm any action before it is taken.
What does the Slack skill make the agent do?
Once installed, the agent can drive a web browser that is already signed into Slack. Typical tasks include checking unread channels and DMs, opening a channel, running a search, reading the messages that are visible on screen, and taking screenshots as evidence.
It works by reading the page's accessibility tree and clicking element references, not by calling the Slack API. Because of this, no Slack tokens, OAuth setup or bot configuration are needed. The actual browsing is done by the agent-browser command-line tool.
- Check unread channels and DMs
- Open channels and navigate the sidebar
- Search Slack and read the results
- Read visible messages and capture screenshots
What you need before installing
You need Node.js with npx, because the skill's allowed tools call agent-browser through Bash. You also need the agent-browser CLI from the vercel-labs/agent-browser project and a Slack session open in a browser.
The browser session can be either attached on remote debugging port 9222, or opened at app.slack.com with you logged in.
- Node.js with npx
- The agent-browser CLI from the vercel-labs/agent-browser project
- A Slack session open in a browser, either on remote debugging port 9222 or at app.slack.com while logged in
How do you install the Slack skill?
Download the skill ZIP from this page and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
If you use the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/
- Claude apps: upload the ZIP under Settings → Capabilities → Skills
~/.claude/skills/How do you connect the browser to Slack?
The skill's quick start recommends attaching to a Slack session that is already open, which is faster than starting a new browser. Start Slack in a browser with remote debugging on port 9222, then connect. If Slack is not running, the skill can open it instead.
After connecting, take a snapshot to see which elements are available. Each interactive element gets a reference such as @e1 or @e2.
agent-browser connect 9222
agent-browser snapshot -i
# Or open Slack if not already running
agent-browser open https://app.slack.comHow do you trigger the skill with natural requests?
You do not need to name commands. The skill is triggered by plain requests that match its description. Phrases the skill lists as triggers include "check my Slack", "what channels have unreads", "send a message to", "search Slack for", "extract from Slack" and "find who said".
Be specific about the goal and the output you want. For example, ask for a list of unread channels with a screenshot, or for the search term and the time period you care about.
- "Check my Slack and tell me which channels have unreads."
- "Search Slack for the release date discussion and screenshot the results."
- "Find who said we should delay the launch in #product-design."
- "Extract the visible messages from the engineering channel into a summary."
How does a typical task run?
The skill follows a repeatable loop: connect or navigate, take a snapshot, click into the right place, read or act, then capture a screenshot. The skill's reference notes that you should re-snapshot after each navigation, because references change when the page changes.
For an unread check, the agent usually looks at the Activity tab, the DMs tab and the "More unreads" button in the sidebar, then records what it found.
- Connect or navigate to Slack
- Snapshot to find element references
- Navigate to the channel, tab or search
- Read the content or perform the action
- Screenshot the result as evidence
agent-browser snapshot -i
agent-browser click @e21 # More unreads button
agent-browser screenshot slack-unreads.pngWhat about safety, scripts and network access?
This skill runs scripts and uses the network. It operates a browser session that is logged in as you, so an agent acting through it can post messages or take other actions with your identity. Review each action before it happens, especially sends and changes.
Screenshots and extracted messages may contain private content and may be shared with whoever receives the agent's output. Do not use the skill in sensitive workspaces where that content must stay in a controlled environment. The skill reads only what is visible on screen, so for complete data it is not a suitable export tool.
- Reading and navigation are the skill's strongest use; the skill covers navigation and reading rather than careful sending
- Check every message before it is posted
- Do not use it where screenshots or message content must not leave a controlled environment
- Slack may rate-limit rapid interactions; add delays if needed
What people use it for
- Get a quick summary of unread channels and DMs without scrolling through every channel.
- Search Slack for a topic and capture screenshots of the results as evidence.
- Pull visible messages from a channel into a weekly status report, using the skill's report template.
- Find out who said something in a channel and record where it appeared.
Frequently asked questions
How do I install the Slack agent skill?
Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills.
What does the Slack skill require?
It requires Node.js with npx, the agent-browser CLI from the vercel-labs/agent-browser project, and a Slack session open in a browser. That session can be attached on remote debugging port 9222 or opened at app.slack.com while you are logged in.
Do I need a Slack API token or bot to use it?
No. The skill uses browser automation rather than the Slack API, so no OAuth setup, webhooks or bot tokens are needed. It reads the page's accessibility tree and clicks element references.
Is the Slack agent skill free to use?
This page does not state pricing or licence terms beyond the listed licence, which is Apache-2.0. Check the source repository for the full licence text before using it in your organisation.
Can the agent send messages for me?
The skill can interact with Slack, but it operates a logged-in browser, so any action can be taken with your identity. Its strongest use is navigation and reading. Review every message before it is sent.
Why can't the skill find a button I can see?
The skill depends on Slack's interface layout, which changes often and can break hard-coded element references such as @e21. Take a fresh snapshot, check that the element appears in the accessibility tree, and scroll if it is off screen.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Choose this skill if you want a quick, screenshot-backed summary of what you missed in Slack, or you already run a browser session with Slack open and want an agent to work inside it. Pick a different tool if you need reliable large-scale export, or if an agent must send messages on your behalf without review, or if the workspace is too sensitive for screenshots and message content to leave a controlled environment.
Download the skill