Check if message should be skipped from rendering. Args: message: The message to check Returns: True if the

Open a streaming message call, print deltas as they arrive, and return the assembled final message (anthropic.types.Message, or BetaMessage when

A server_tool_use block as the SDK would return it (an object, not a dict).

Best-effort name of the GUI app this process is running under, so the user knows which row to enable in the permission list.

Layout-independent typing. pyautogui.write() sends hardcoded ANSI virtual keycodes (physical key positions on a US QWERTY board), which macOS

Wait for a specified duration.

Write content to a file.

Take a screenshot of a specific region.

Pre-tool-use hook that validates bash commands using an allowlist. Only commands in ALLOWED_COMMANDS are permitted. Args: input

Call a tool on the MCP server with provided arguments.

Copy the app spec file into the project directory for the agent to read.

Count passing and total tests in feature_list.json. Args: project_dir: Directory containing feature_list.json Returns:

Create a Claude Agent SDK client with multi-layered security. Args: project_dir: Directory for the project model: Claude mod

Factory function to create the appropriate MCP connection.

Create a ZIP archive containing the transcript and optionally images. Args: messages: List of message dictionaries from session state

Execute the tool with provided parameters.

Execute multiple tools sequentially or in parallel.

Extract images from messages and create transcript with file references. Returns: Tuple of (transcript_json, list_of_base64_image_data)

Format messages for Claude API with optional caching.

Load the coding agent prompt.

Find the specific command segment that contains the given command. Args: cmd: The command name to find segments: List of com

Get the documented size for the given viewport's aspect ratio. Args: viewport_width: Actual viewport width v

Load the initializer prompt.

Retrieve available tools from the MCP server.

Main application entry point.

Render loop for streamlit

Main entry point.

Run the test suite.

Parse command line arguments.

Print a formatted header for the session.

Process content using appropriate processors.

Process text content type.

Run the browser automation agent with user input.

Run a single agent session using Claude Agent SDK. Args: client: Claude SDK client message: The prompt to send proje

Run all test cases.

Run agent with MCP tools asynchronously.

Run the autonomous agent loop. Args: project_dir: Directory for the project model: Claude model to use max_iteration

Run the agent. Each user message gets up to ``max_iters`` model turns. When ``interactive`` is true, the loop prompts for a follow-up after the

Set up MCP server connections and create tool interfaces.

Split a compound command into individual command segments. Handles command chaining (&&, ||, ;) but not pipes (those are single commands).

Implement the str_replace command, which replaces old_str with new_str in the file content

Test the command extraction logic.

Test chmod command validation.

Test init.sh script execution validation.

Convert tool to Claude API format.

Convert all tools to API parameters.

Remove oldest messages when context window limit is exceeded.

Validate required environment variables are set.

Validate pkill commands - only allow killing dev-related processes. Uses shlex to parse the command, avoiding regex bypass vulnerabilities.

Implement the view command