- Updated
- Version
- v1.1
- Sources
- 1 source
- Status
- Published
browser-use
Browser Use automation
Browser Use
Controls a browser for a bounded objective, inspects accessible state, and makes external effects explicit.
- License
- MIT
- Risk
- Risk level: high
Risk profile
This is a risk profile, not a certification or a safety guarantee.
What can it read?
Present
- It can read the content of web pages it opens, including what is visible in the browser.
What can it create or change?
Present
- It can click, fill forms, and save screenshots or files. It can also change an account or make a purchase when that is asked.
Can it use the internet or send data?
Present
- It can use the internet through a browser, including a remote cloud browser session.
Can it use saved logins or a paid quota?
Present
- It can use a paid cloud-browser account and, with your go-ahead, saved browser logins.
Which action needs your approval?
Present
- Using a signed-in browser session, sending data outside, changing an account, or making a purchase needs your explicit approval first.
When was the source verified, and has the snapshot changed since?
Present
- The source was last verified on 2026-07-19. The published snapshot still matches that verification. This is not a live check of the upstream source.
- Canonical source
- https://github.com/browser-use/browser-use/blob/950eb03617e67548d759c02beac1ad122c6b6458/skills/browser-use/SKILL.md(opens in a new tab)
- Source revision
- 950eb03617e67548d759c02beac1ad122c6b6458
- License
- MIT
- Verification date
- 2026-07-19
- Source status
- Verified
- Change status
- Snapshot consistent
- The published snapshot still matches the last verification. This is not a live check of the upstream source.
Understand
Covered goals
- navigate a website
- inspect accessible page state
- automate a browser flow
Inputs → outputs
- target URL and bounded browser objective
- local Chrome/CDP or Browser Use Cloud execution mode
- explicit authorization for authenticated state and external mutations
- browser observations and inspected accessible or DOM state
- interaction results and optional screenshots, recordings, downloads, or other files
Recognize
Verbatim from the English source
Triggers
- open and inspect a website
- fill or test a browser flow
- capture browser evidence
Non-triggers
- server-side HTTP request only
- bypass access controls
- silent production mutation
Choose
Verbatim from the English source
Permissions
- launch or connect to local or remote browser sessions
- read rendered page content
- access existing cookies, profiles, or SSO state only when explicitly authorized
- interact with page controls and create captures or files when authorized
External effects
- network requests
- form submissions, messages, purchases, or account changes when explicitly requested
- remote Browser Use Cloud sessions can incur charges until stopped or timed out
- screenshots, recordings, downloads, and other browser artifacts can be written to disk
- authenticated cookies or profiles can be synchronized to a cloud browser when explicitly authorized
Compatibility
- an environment that can run the browser-use CLI and Python snippets against local CDP or Browser Use Cloud
Required tools
- browser-use CLI
- local Chrome or Chromium with CDP remote debugging, or an optional Browser Use Cloud browser
- Browser Use API key and authenticated CLI session for cloud mode
Check permissions and confirm every external effect before using this skill.
Compare
None declared
Understand → Recognize → Choose → Compare
Pack for your agent
Pre-written instruction by SkillCodex — your request is neither sent nor used to adapt this text; no content is generated, and copying executes nothing.
Use in your agent
Complete the task with Browser Use automation
# Use Browser Use automation in your agent ## Objective Use the existing browser-use skill only when it directly covers the requested task. ## Prerequisites - target URL and bounded browser objective - local Chrome/CDP or Browser Use Cloud execution mode - explicit authorization for authenticated state and external mutations - browser-use CLI - local Chrome or Chromium with CDP remote debugging, or an optional Browser Use Cloud browser - Browser Use API key and authenticated CLI session for cloud mode ## Permissions - launch or connect to local or remote browser sessions - read rendered page content - access existing cookies, profiles, or SSO state only when explicitly authorized - interact with page controls and create captures or files when authorized ## External effects - network requests - form submissions, messages, purchases, or account changes when explicitly requested - remote Browser Use Cloud sessions can incur charges until stopped or timed out - screenshots, recordings, downloads, and other browser artifacts can be written to disk - authenticated cookies or profiles can be synchronized to a cloud browser when explicitly authorized ## Procedure - Open the canonical source pinned to revision 950eb03617e67548d759c02beac1ad122c6b6458. - Read the skill instructions before calling its tools. - Ask for confirmation before any paid, destructive, or remote effect not already authorized. - Run the procedure within the requested scope and retain useful evidence. ## When not to use - server-side HTTP request only - bypass access controls - silent production mutation ## Expected result - browser observations and inspected accessible or DOM state - interaction results and optional screenshots, recordings, downloads, or other files ## Guardrails - Show the proposed changes before any external action. - Do not publish, send, delete, pay for, or change remote state without explicit authorization. - Preserve unrelated changes and stop if the scope becomes ambiguous. ## Output format - Outcome or verdict. - Files or actions involved. - Checks run and observable evidence. - Remaining blockers or limitations.
- Requires · browser-use CLI
- Requires · local Chrome or Chromium with CDP remote debugging, or an optional Browser Use Cloud browser
- Requires · Browser Use API key and authenticated CLI session for cloud mode
Why it works
- Activation is limited to tasks the skill actually covers.
- Permissions, tools, and external effects are visible before execution.
- Confirmation and the output format keep actions controllable.
Try next
Check the Browser Use automation result
# Verify the Browser Use automation outcome ## Objective Check that the browser-use skill was used within scope and that its effects match the authorized request. ## Checks - Compare the source and revision with the manifest. - Separate local, browser, and external evidence. - Report any permission or effect that was not disclosed. ## Guardrails - Show the proposed changes before any external action. - Do not publish, send, delete, pay for, or change remote state without explicit authorization. - Preserve unrelated changes and stop if the scope becomes ambiguous. ## Output format - Outcome or verdict. - Files or actions involved. - Checks run and observable evidence. - Remaining blockers or limitations.
sha256:9b004245c244db6fe11e4b8bd423afd09b9c6bfa62494d4b63f71bd42319a89c
Verify in your agent
Confirm that Browser Use automation produced the right outcome
# Verify Browser Use automation in your agent ## Result to check Verify the use of browser-use for the current bounded task. ## Failure signals - The source or revision does not match the manifest. - A permission or external effect was not disclosed. - The output does not match the expected artifacts. - The skill activated for a declared non-trigger. ## Checks - Compare the SHA-256 of SKILL.md with sha256:1ae1697ca45c8604110d1b80274a88f9736e45466f81ac15f38954e5ea7efb2b. - Confirm the permissions actually used and the observed effects. - Verify artifacts with the narrowest available test. - Separate local, browser, and external evidence in the report. ## Expected result - browser observations and inspected accessible or DOM state - interaction results and optional screenshots, recordings, downloads, or other files ## Guardrails - Show the proposed changes before any external action. - Do not publish, send, delete, pay for, or change remote state without explicit authorization. - Preserve unrelated changes and stop if the scope becomes ambiguous. ## Output format - Outcome or verdict. - Files or actions involved. - Checks run and observable evidence. - Remaining blockers or limitations.
- Requires · browser-use CLI
- Requires · local Chrome or Chromium with CDP remote debugging, or an optional Browser Use Cloud browser
- Requires · Browser Use API key and authenticated CLI session for cloud mode
Why it works
- The check starts from the manifest and expected artifacts, not a general impression.
- Observed permissions and effects are compared with what was disclosed.
- Local, browser, and external evidence stay separate in the verdict.
Try next
Prepare the next Browser Use automation check
# Make the Browser Use automation check reusable ## Objective Turn the successful checks into a bounded checklist for the next use of browser-use. ## Checks - Keep only observable signals. - Tie each permission to its expected effect. - Add an explicit stop if the source or revision changes. ## Guardrails - Show the proposed changes before any external action. - Do not publish, send, delete, pay for, or change remote state without explicit authorization. - Preserve unrelated changes and stop if the scope becomes ambiguous. ## Output format - Outcome or verdict. - Files or actions involved. - Checks run and observable evidence. - Remaining blockers or limitations.
sha256:65ba2272982b6313c4f7fbe7b46d8df2dc8bbd0b1b541dd101e54e6a66d7781d
Pack digest: sha256:52e6431829a17b291b820305dc486a298cec8c481a8413bc984651ee441cea4f
Source provenance
Open the canonical source (opens in a new tab)- Source revision
- 950eb03617e67548d759c02beac1ad122c6b6458
- Digest
- sha256:1ae1697ca45c8604110d1b80274a88f9736e45466f81ac15f38954e5ea7efb2b