Skip to content

Default actions

This page lists every default (built-in, non-integration) action with its action set, parameters, and purpose. It is generated from the @action(...) decorators in CraftBot's app/data/action/ by scripts/gen_default_actions.py, so it reflects the code as of the last time that script was run. For how actions are registered, grouped into sets, selected, and executed, read Actions and action sets.

Two other groups of actions exist beyond this page. Integration actions (for example create_github_issue) are documented on each service's page in Integrations. MCP tool actions are created dynamically as mcp_<server>_<tool> when you connect an MCP server.

The Set column shows which action set the action belongs to. Actions in core are available in every task. Actions in other sets are available only when the task selected that set. Actions marked Irreversible are guarded so a retried turn does not repeat them.

There are 64 default actions.

Task control

Action Set Parameters What it does
set_requirement core requirements Record (or update) the concrete, checkable requirements that define DONE for the current deliverable.
send_message core message, continue_work Use this action to deliver a text update to the user; it is recorded in the conversation log and event stream. Irreversible.
send_message_with_attachment core message, file_paths, continue_work Send a message to the user with one or more file attachments. Irreversible.
end_turn core none End the current run without sending any message.
wait core seconds Pause execution for a specified duration.
spawn_subagent core none See the source for details.
sub_task_end n/a status, result End your sub-agent run.
Action Set Parameters What it does
read_file core file_path, encoding, offset, limit, max_line_length Reads a file and returns its contents with line numbers.
write_file core file_path, content, encoding, mode Write or overwrite a text file with the provided content.
stream_edit core file_path, old_string, new_string, replace_all, regex, ignore_case Performs string replacement in a file.
find_files core pattern, recursive, base_directory, all_drives, limit Finds files by name or pattern across the system.
grep_files core none Searches files for a regex pattern and returns results.
list_folder core path Lists the contents of a specified folder/directory.

Web research

Action Set Parameters What it does
web_search core query, num_results Performs web search and returns search result snippets with markdown hyperlinks.
web_fetch core url, mode, timeout, max_content_length, use_jina_fallback Fetches a URL and returns cleaned text/markdown content.

Shell and HTTP

Action Set Parameters What it does
run_shell core command, shell, timeout, cwd, env, background Executes a shell command using the appropriate OS shell, capturing stdout, stderr, and exit code.
http_request core method, url, headers, params, json, data, timeout, allow_redirects, verify_tls, save_to, include_headers Sends HTTP requests (GET, POST, PUT, PATCH, DELETE) with optional headers, params, and body.

Clipboard

Action Set Parameters What it does
clipboard_read core none Read the current content from the system clipboard.
clipboard_write core content Write text content to the system clipboard.

Scheduling

Action Set Parameters What it does
schedule_task scheduler, core, proactive name, instruction, schedule, priority, enabled, action_sets, skills, payload Schedule a task for execution.
scheduled_task_list scheduler, core, proactive none List all scheduled tasks configured in the scheduler.
schedule_task_toggle scheduler, proactive schedule_id, enabled Enable or disable a scheduled task by its ID.
remove_scheduled_task scheduler, core, proactive schedule_id Remove a scheduled task from the scheduler by its ID.

Recurring proactive tasks

Action Set Parameters What it does
recurring_add proactive name, frequency, instruction, time, day, priority, permission_tier, enabled, conditions Add a new recurring task to PROACTIVE.md.
recurring_read proactive frequency, enabled_only Read and parse recurring tasks from PROACTIVE.md.
recurring_update_task proactive task_id, updates, add_outcome Update an existing recurring task in PROACTIVE.md.
recurring_remove proactive task_id Remove a recurring task from PROACTIVE.md.

Memory

Action Set Parameters What it does
memory_search core none Search the agent's memory for relevant information based on a semantic query.

Skills and action sets

Action Set Parameters What it does
list_skills core none List all enabled skills with their names and descriptions.
use_skill core skill_name Load a skill into this session: its instructions are injected into your context and its recommended action sets are loaded.
list_action_sets core none List all available action sets and their descriptions.
add_action_sets core action_sets Load additional action sets from the capability catalog to expand the actions available in this session.
remove_action_sets core action_sets Unload action sets from this session to reduce available actions.

Integration management

Action Set Parameters What it does
list_available_integrations core filter_connected List all available external app integrations and their connection status.
connect_integration core integration_id, credentials, auth_method Connect an external app integration.
check_integration_status core integration_id, session_id Check the connection status of a specific integration, or check the status of an ongoing WhatsApp QR code session.
disconnect_integration core integration_id, account_id Disconnect an external app integration.

Documents and PDF

Action Set Parameters What it does
read_pdf document_processing file_path, mode, page_range, include_metadata Reads a PDF and returns its content.
edit_pdf document_processing file_path, output_path, operations Edits an existing PDF by applying a list of operations and saves the result to a new file.
convert_to_pdf document_processing none Universal source-to-PDF converter.
convert_from_pdf document_processing source_path, output_path, target_format, mode Universal PDF-to-source converter.
convert_to_markdown document_processing input_file, output_md Cleans scraped text from .txt, .md, or .docx and converts it into clean, well-structured Markdown suitable for PDF conversion.
perform_ocr document_processing, image, video image_path, user_prompt Extracts all text from an image using OCR via a Vision Language Model.

Media understanding and generation

Action Set Parameters What it does
describe_image core, document_processing, image image_path, prompt Uses a Visual Language Model to analyse an image and return a detailed, markdown-ready description.
understand_video document_processing, image, video video_path, query, max_frames Uses the configured VLM model (default: Gemini 1.5 Pro) for native video understanding when a Google API key is configured.
generate_image content_creation, image, document_processing prompt, output_path, resolution, aspect_ratio, number_of_images, negative_prompt, reference_images, safety_filter_level Generates an image from a text prompt using the configured image generation provider (OpenAI gpt-image-2 or Google Gemini).
generate_video content_creation, video prompt, output_path, duration_seconds, aspect_ratio, resolution, number_of_videos, negative_prompt, reference_image, last_frame, reference_images, seed, with_audio, person_generation, camera_fixed, watermark, callback_url Generates a video from a text prompt using the configured video generation provider (Google Gemini Veo, OpenAI Sora, or BytePlus Seedance).

Living UI

Action Set Parameters What it does
living_ui_scaffold living_ui name, description, features, chat_context, theme, auth_mode Create and register a new Living UI project from the template, then dispatch the build to the project's dedicated session.
living_ui_notify_ready living_ui project_id Launch or RELAUNCH a Living UI project: installs dependencies, runs the validation gate, restarts backend and frontend, notifies the browser.
living_ui_restart living_ui project_id Restart a Living UI project (backend + frontend).
living_ui_report_progress living_ui project_id, phase, progress, message Report progress ONLY during Living UI creation (initial build).
living_ui_import_zip living_ui zip_path, name Import a Living UI project from an exported ZIP file (round-trip with export): registers it as a NEW project with fresh identity and port, strips shipped credentials, and re-vendors the kit. Irreversible.
living_ui_http living_ui project_id, method, path, headers, params, json, data, timeout, target FALLBACK ONLY — prefer the lui CLI via run_shell (node /living-ui/tools/src/cli.ts ops

Other actions

Actions present in the code that are not yet grouped above.

Action Set Parameters What it does
browser_probe living_ui url, steps, project_path Drive a RUNNING Living UI in a headless browser (invisible — no window).
living_ui_approve_triggers living_ui project_id Approve a Living UI app's declared agent triggers (its triggers.json — requests the app may fire at the agent).
living_ui_convert living_ui source, name, description REBUILD a foreign (non-Living-UI) app as a Living UI: scaffolds a fresh Living UI project, ships the original source (zip / folder / git URL) as read-only reference material, synthesizes the requireme... Irreversible.
living_ui_import living_ui source, name Import a Living UI project from ANY source: an exported ZIP file, a local folder path, or a git URL (GitHub downloads fast; other hosts are cloned depth-1). Irreversible.
living_ui_list_projects living_ui none List the user's Living UI projects: id, name, status, URL, path, and whether the app is delivered.
living_ui_marketplace_install living_ui app_id, name, description, will_adapt Install a pre-built Living UI app from the marketplace by id (resolve ids with living_ui_marketplace_list). Irreversible.
living_ui_marketplace_list living_ui none List the Living UI marketplace catalogue: pre-built apps the user can install by id.
living_ui_usage living_ui project_id Get the operating manual for a Living UI project: its path, data schema, and the exact lui CLI commands (run via run_shell) to read/write its data and run its operations.
living_ui_walk_verify living_ui project_id Run the independent walk-verify sub-agent against the RUNNING Living UI project: a real browser (headless) drives the app feature-by-feature against reference/requirements.md.
unload_skill core skill_name Unload a previously loaded skill from this session, removing its instructions from your context.
update_todos core todos Update the todo list for the current run of this session.

Next