ai-dev-swarm

To click a page element in the browser, perform a click on buttons, links, or controls during Playwright automation.

To close the current Playwright page, close the page to end the session or free resources.

To read console logs from the current page, retrieve console messages for debugging JavaScript output.

To drag and drop between elements on a web page, perform a drag interaction from a source to a target.

To run JavaScript in the page context, evaluate an expression on the page or a specific element for inspection or manipulation.

To upload files through a file input, provide one or more file paths for a page file chooser.

To fill multiple form fields at once, set values for inputs, checkboxes, and selects in a single call.

To accept or dismiss browser dialogs like alert/confirm/prompt, handle the dialog so automation can continue.

To hover over a page element, trigger hover states such as menus or tooltips.

To install the Playwright browser defined in config, run this when the browser is missing or not installed.

To go back in browser history, navigate to the previous page from the current page.

open a URL, open a web page, open web browser, or navigate the current page in the web browser to a new URL.

To inspect network activity since page load, list network requests to review API calls and resources.

To press a keyboard key in the page, send a key press for shortcuts or input behavior.

To resize the browser window, change viewport size to test responsive layouts.

To run a custom Playwright script on the page, execute a code snippet for advanced interactions.

To choose an option in a dropdown, select one or more values in a select element.

To capture an accessibility snapshot of the current page, take a structured snapshot for UI inspection.

To manage browser tabs, list, create, select, or close Playwright tabs.

To take a screenshot of a page or element, capture an image for visual verification.

To type text into an editable field, enter text into inputs or textareas.

To wait for page state changes, wait for text to appear or disappear or for a timeout.

Draft a conventional commit message when the user asks to commit, commit changes, git commit, or save changes to git. Use when user says "commit", "commit this", "commit changes", "make a commit", or similar.

Perform git partial merges (checkout specific folders from another branch) and squash merges (clean single-commit integration) to selectively integrate changes across branches.

Add and manage Model Context Protocol (MCP) servers for AI agents, including transports, scopes, and config files. Use when extending agent capabilities with new tools or data sources.

To run an exact AWS CLI command, execute a validated `aws ...` command for direct AWS operations when you already know the precise service and parameters; use instead of suggesting commands.

To get possible AWS CLI commands from a natural-language request, suggest candidate `aws ...` commands when the exact service, operation, or syntax is unclear; use before call-aws.

To remove a finished background process from the manager list, clear a stopped process after it exits so it no longer appears in listings.

To read logs from a managed background process, fetch recent output (head or tail) for a process so you can inspect its stdout/stderr.

To check if the Background Process Manager is running, get the server status and confirm it is available.

To see all managed background jobs, list processes so you can find their IDs, states, and names.

To run a short command and wait for the result, execute a synchronous shell command via the manager and return full output.

To start a long-running background task, launch a managed process such as a server or watcher and keep it tracked.

To stop a managed background task, terminate a running process by ID to shut it down cleanly.

To fetch up-to-date docs and code examples from Context7 for a known library, query by a resolved library ID after running resolve-library-id (max 3 calls).

To find the Context7-compatible library ID for a package or product name, resolve the library ID before querying docs unless the user already provided an /org/project ID (max 3 calls).

To register project roots for Dart tooling access, add one or more root paths before using other Dart tools on those projects.

To run static analysis across a Dart or Flutter project, analyze files to find compile and lint errors.

To connect to the Dart Tooling Daemon for editor/runtime data, connect using a user-provided DTD URI before using related Dart tools.

To scaffold a new Dart or Flutter project, create a fresh project structure and starter files.