built-in-browser-automation

name	built-in-browser-automation
description	Drive browser tabs with dispatchAction for navigation, tab orchestration, snapshotting, element queries/interactions, visual capture, cookies/storage, debugger CDP access, and lightweight session recording. Use when tasks require direct browser automation in the current Chrome tab context.

Built-in Browser Automation Skills

This is the root index for all built-in browser automation capabilities. Use this skill for any task that requires interacting with web pages, controlling browser tabs, or automating web workflows.

Core Capabilities

Category	Description	Reference File
Browser Orchestration	Navigate pages, read page info, manage tabs, and check debugger connectivity	references/browser-orchestration.md
Snapshot & Reference Management	Capture page structure, resolve element references, manage tab state	references/snapshot-refs.md
Element Query & Read Operations	Extract text, attributes, state, and properties from DOM elements	references/element-queries.md
Element Interaction	Click, type, fill forms, scroll, and manipulate page elements	references/element-interactions.md
Visual Operations	Capture screenshots and visual evidence	references/visual-operations.md
Cookie Management	Read, set, and clear browser cookies	references/cookie-management.md
Storage Management	Access and modify localStorage, sessionStorage	references/storage-management.md
Debugger & CDP	Attach debugger and send raw Chrome DevTools Protocol commands	references/debugger-cdp.md
Session Recording	Record and replay browser interaction sessions	references/recording.md

Preconditions

1. Browser Tab ID: Ensure typeId is set to the target browser tab ID
2. Navigation First: Prefer NAVIGATE for page loads instead of raw SEND_CDP calls
3. Element References: Always prefer selectorOrRef values from a recent GET_SNAPSHOT result
4. CDP Risk: Treat SEND_CDP as high-risk - only use when standard actions cannot solve the task

Recommended Workflow

flowchart LR
    A[Start] --> B[NAVIGATE or TAB_SWITCH]
    B --> C[GET_SNAPSHOT to capture page state]
    C --> D[Query element state/attributes]
    D --> E[Execute interactions]
    E --> F[GET_SNAPSHOT to verify changes]
    F --> G{Task complete?}
    G -->|No| D
    G -->|Yes| H[TAKE_SCREENSHOT if evidence needed]
    H --> I[End]

Guardrails & Best Practices

• Minimal Commands: Keep automation sequences minimal and deterministic
• High-level First: Prefer dedicated actions (NAVIGATE, TAB_*, GET_*) before falling back to SEND_CDP
• Ref Validity: Always re-snapshot after page navigation or significant content changes before reusing old element references
• CDP Usage: Avoid raw SEND_CDP commands unless absolutely necessary - prefer higher-level actions
• Error Handling: Check action results and validate state changes after each operation
• Performance: Batch related operations where possible to reduce round-trips

Additional Resources

For detailed usage examples and parameter specifications, refer to the individual category documentation files linked above.

• Machine-readable action-to-doc index: references/action-index.json