[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Complete task. Always call this once you're finished with a browser task, whether it succeeded or not — besides signaling completion, on success this automatically saves everything you did since the last find_recipe/done as a recipe (using text as its instruction), so a future task on this site can replay it via find_recipe instead of redoing every step. This is the only thing that triggers auto-saving — individual actions (go_to_url, click_element, etc.) are silently recorded as you go, but nothing is persisted until done() is called with success: true.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Search the query in Google. Opens a new tab for this task if none is active yet, otherwise reuses the tab already in use for this task — never the user's own currently-focused tab. The query should be a search query like humans search in Google, concrete and not vague or super long. More the single most important items.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Navigate to URL. Opens a new tab for this task if none is active yet, otherwise navigates the tab already in use for this task — never the user's own currently-focused tab, and every later action (click_element, read_page, etc.) keeps targeting that same tab even if the user clicks into a different one meanwhile.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Go back to the previous page

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Wait for x seconds default 3, do NOT use this action unless user asks to wait explicitly

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Read the current page in this task's tab: URL, title, and an indexed list of interactive elements. Call this before click_element, input_text, or the dropdown tools to get valid indices. Body text is omitted by default — pass include_text when you actually need to read page content, not just interact with it.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Click element by index

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Input text into an interactive input element

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Switch to tab by tab id, and make it the tab this task uses from now on (e.g. after a link opened a new tab).

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Open URL in a new tab, and use it as this task's tab from now on.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Close tab by tab id

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Cache what you have found so far from the current page for future use

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Scrolls to a particular vertical percentage of the document or an element. If no index of element is specified, scroll the whole document.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Scroll the document in the window or an element to the top

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Scroll the document in the window or an element to the bottom

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Scroll the document in the window or an element to the previous page. If no index is specified, scroll the whole document.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Scroll the document in the window or an element to the next page. If no index is specified, scroll the whole document.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] If you dont find something which you want to interact with in current viewport, try to scroll to it

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Send strings of special keys like Backspace, Insert, PageDown, Delete, Enter. Shortcuts such as Control+o, Control+Shift+T are supported as well. This gets used in keyboard press. Be aware of different operating systems and their shortcuts

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Get all options from a native dropdown

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Select dropdown option for interactive element index by the text of the option you want to select

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Execute arbitrary JavaScript in this task's tab (page context) and return the result.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] List every previously saved step-by-step trajectory for this domain, before doing the task manually. Returns each saved recipe's own instruction text plus its fingerprint and success/failure counts — no steps yet. Read the instructions like retrieved documents and judge for yourself whether any of them is the same task you're about to do, even if worded differently ('post a tweet' vs 'tweet something'); ignore recipes that don't actually match. If one fits, pass its fingerprint straight into run_recipe to replay the whole task in one call instead of one tool call per step. Always try this first for any task that sounds like something you (or a prior session) may have already done on this site.

[Real Chrome browser via the OctoLimb extension — drive the user's browser; use only when the task requires acting on a real website.] Execute a sequence of OctoLimb tool steps in one call, without round-tripping to you between each one — use this to batch obvious action chains (e.g. go_to_url, then click_element, then input_text) or to replay steps returned by find_recipe. Runs steps exactly as given, in order, via the real browser — nothing is added or skipped, so include every step you'd need if doing this by hand one call at a time (e.g. a read_page before an index-based step on a page you haven't indexed yet in this session). Stops at the first step that errors, returning every result so far plus which step failed — pick up manually from there. Pass fingerprint (from find_recipe) to replay and auto-update its success/failure count instead of passing steps directly. Pass save_as alongside steps to save this task once it fully succeeds — this saves every tool call made in the session so far (including ones made outside this run_recipe call, e.g. an earlier read_page), not just the steps in this call, so replay later needs nothing extra.