name: skyvern-browser-automation description: "AI-powered browser automation — navigate sites, fill forms, extract structured data, log in with stored credentials, and build reusable workflows." category: browser-automation risk: safe source: community source_repo: Skyvern-AI/skyvern source_type: official date_added: "2026-04-23" author: mark1ian tags: [browser-automation, mcp, web-scraping, form-filling, ai-agents, workflow-automation] tools: [claude, cursor, gemini, codex] license: "AGPL-3.0" license_source: "https://github.com/Skyvern-AI/skyvern/blob/main/LICENSE"

Skyvern Browser Automation -- CLI Judgment Procedure

Skyvern uses AI to navigate and interact with websites. Every command below is a runnable skyvern <command> invocation.

When to Use This Skill

• Use when you need AI-assisted browser automation for navigation, extraction, form filling, login flows, or reusable website workflows.
• Use when deterministic selectors are unavailable and Skyvern's visual/a11y reasoning can identify page controls.
• Use when a one-off browser task should become a repeatable workflow with run history and verification.

Step 1: Classify Your Task (ALWAYS do this first)

MCP note: if you are using the Skyvern MCP instead of the CLI, prefer observe + execute for same-page multi-step UI work. The CLI does not expose that pair directly.

Step 2: Apply These Decision Rules

1. If the prompt includes a selector, id, XPath, or exact field target, use browser primitives -- not act.
2. If you only need a yes/no answer, use validate -- not extract or act.
3. If the work stays on one page and labels are clear, use act or a primitive chain.
4. If the user says try this once, see if this works, or clearly wants a one-off exploratory trial, use run-task.
5. If the task spans multiple pages and is meant to be reusable, scheduled, repeatable, or explicitly set up as automation, use workflow create.
6. Never type passwords. Always use stored credentials with skyvern browser login.

Step 3: Create a Session

Every browser command needs a session. Create one first:

# Cloud session (default -- works for public URLs)
skyvern browser session create --timeout 30

# Local session (for localhost URLs or self-hosted mode)
skyvern browser session create --local --timeout 30

# Connect to existing browser via CDP
skyvern browser session connect --cdp "ws://localhost:9222"

Session state persists between commands. After session create, subsequent commands auto-attach. Override with --session pbs_.... Close when done: skyvern browser session close.

Step 4: Execute by Classification

Quick check (yes/no)

skyvern browser validate --prompt "Is the user logged in? Look for a dashboard or avatar."

Returns true/false. Cheapest AI option -- prefer over extract or act for boolean checks.

Quick inspection

skyvern browser extract \
  --prompt "Extract all product names and prices" \
  --schema '{"type":"object","properties":{"items":{"type":"array","items":{"type":"object","properties":{"name":{"type":"string"},"price":{"type":"string"}}}}}}'

Uses screenshots + dedicated extraction LLM. Better than screenshot+read because Skyvern's LLM interprets the page.

Single action (known target)

skyvern browser click --selector "#submit-btn"
skyvern browser type --text "user@co.com" --selector "#email"
skyvern browser select --value "US" --intent "the country dropdown"

Deterministic. No AI. Three targeting modes:

1. Intent: --intent "the Submit button" (AI finds element)
2. Selector: --selector "#submit-btn" (CSS/XPath, deterministic)
3. Hybrid: both (selector narrows, AI confirms)

Single action (unknown target)

skyvern browser act --prompt "Click the Sign In button"
skyvern browser act --prompt "Close the cookie banner, then click Sign In"

Warning: act has NO screenshots in its LLM reasoning. It uses an economy accessibility tree. Fine for well-labeled elements. For visually complex targets, use MCP observe+click or hybrid mode.

Same-page multi-step

skyvern browser act --prompt "Fill the shipping form and click Continue"

Use act when the fields and buttons are clearly labeled and the flow stays on one page. If you need tighter control, break the work into click, type, select, press-key, and wait.

Throwaway autonomous trial

skyvern browser run-task \
  --url "https://example.com" \
  --prompt "Check whether the checkout flow works end to end and extract the confirmation number"

Use run-task to prove feasibility or do one-off exploration. If the task becomes important enough to rerun, debug, or share, convert it to a workflow.

Multi-page or reusable automation — build a workflow with one block per step

skyvern workflow create --definition @checkout-workflow.yaml
skyvern workflow run --id wpid_123 --wait
skyvern workflow status --run-id wr_789

Each navigation block runs with visual reasoning + verification. Split complex flows into multiple blocks (one per page/step). First run uses AI; subsequent runs replay cached scripts.

Repeated/production

skyvern workflow create --definition @workflow.yaml
skyvern workflow run --id wpid_123 --params '{"email":"user@co.com"}'
skyvern workflow status --run-id wr_789

Split into one block per step. Use navigation blocks for actions, extraction for data. First run uses AI; subsequent runs replay a cached script (10-100x faster). Set --run-with agent to force AI mode for debugging.

Step 5: Verify

Always verify after page-changing actions:

skyvern browser screenshot                          # visual check
skyvern browser validate --prompt "Was the form submitted successfully?"  # boolean assertion
skyvern browser evaluate --expression "document.title"                    # JS state check

Step 6: Error Recovery

Credentials

NEVER type passwords through skyvern browser type or act. Always use stored credentials:

skyvern credentials add --name "my-login" --type password --username "user@co.com"
skyvern credential list                          # find the credential ID
skyvern browser login --url "https://login.example.com" --credential-id cred_123

Types: password, credit_card, secret. Also supports bitwarden, 1password, and azure_vault providers.

Workflow Quick Reference

skyvern workflow create --definition @workflow.yaml   # create
skyvern workflow run --id wpid_123 --wait             # run and wait
skyvern workflow status --run-id wr_789               # check status
skyvern workflow list --search "invoice"              # find workflows
skyvern block schema --type navigation                # discover block types
skyvern block validate --block-json @block.json       # validate before creating

Engine: known path = 1.0 (default). Dynamic planning = 2.0. Split into multiple 1.0 blocks when in doubt. Status lifecycle: created -> queued -> running -> completed | failed | canceled | terminated | timed_out

Common Patterns

Login flow:

skyvern credential list                          # find credential ID
skyvern browser session create
skyvern browser navigate --url "https://login.example.com"
skyvern browser login --url "https://login.example.com" --credential-id cred_123
skyvern browser validate --prompt "Is the user logged in?"
skyvern browser screenshot

Pagination loop:

skyvern browser extract --prompt "Extract all rows"
skyvern browser validate --prompt "Is there a Next button that is not disabled?"
# If true:
skyvern browser act --prompt "Click the Next page button"
# Repeat extraction. Stop when: no next button, duplicate first row, or max page limit.

Debugging:

skyvern browser screenshot                       # visual state
skyvern browser evaluate --expression "document.title"
skyvern browser evaluate --expression "document.querySelectorAll('table tr').length"

Limitations

• Do not use Skyvern to bypass site access controls, rate limits, consent gates, or terms that prohibit automation.
• Browser automation can change remote state; confirm user intent before submitting forms, purchasing, deleting, or sending messages.
• Prefer deterministic selectors for stable production flows; AI actions can misread unlabeled or visually ambiguous controls.
• Store credentials only in the supported credential vaults and never type passwords directly through type or act.

Agent Mode

All commands accept --json for structured output. Set SKYVERN_NON_INTERACTIVE=1 to prevent prompts. Use skyvern capabilities --json for full command discovery. See references/agent-mode.md.

Deep-Dive References