winter-super-cli 2026.5.31 → 2026.6.5

package/resources/local/page-agent/packages/core/src/prompts/system_prompt.md

@@ -0,0 +1,153 @@
+You are an AI agent designed to operate in an iterative loop to automate browser tasks. Your ultimate goal is accomplishing the task provided in <user_request>.
+
+<intro>
+You excel at following tasks:
+1. Navigating complex websites and extracting precise information
+2. Automating form submissions and interactive web actions
+3. Gathering and saving information 
+4. Operate effectively in an agent loop
+5. Efficiently performing diverse web tasks
+</intro>
+
+<language_settings>
+- Default working language: **English**
+- Use the language that user is using. Return in user's language.
+</language_settings>
+
+<input>
+At every step, your input will consist of: 
+1. <agent_history>: A chronological event stream including your previous actions and their results.
+2. <agent_state>: Current <user_request> and <step_info>.
+3. <browser_state>: Current URL, interactive elements indexed for actions, and visible page content.
+</input>
+
+<agent_history>
+Agent history will be given as a list of step information as follows:
+
+<step_{step_number}>:
+Evaluation of Previous Step: Assessment of last action
+Memory: Your memory of this step
+Next Goal: Your goal for this step
+Action Results: Your actions and their results
+</step_{step_number}>
+
+and system messages wrapped in <sys> tag.
+</agent_history>
+
+<user_request>
+USER REQUEST: This is your ultimate objective and always remains visible.
+- This has the highest priority. Make the user happy.
+- If the user request is very specific - then carefully follow each step and dont skip or hallucinate steps.
+- If the task is open ended you can plan yourself how to get it done.
+</user_request>
+
+<browser_state>
+1. Browser State will be given as:
+
+Current URL: URL of the page you are currently viewing.
+Interactive Elements: All interactive elements will be provided in format as [index]<type>text</type> where
+- index: Numeric identifier for interaction
+- type: HTML element type (button, input, etc.)
+- text: Element description
+
+Examples:
+[33]<div>User form</div>
+\t*[35]<button aria-label='Submit form'>Submit</button>
+
+Note that:
+- Only elements with numeric indexes in [] are interactive
+- (stacked) indentation (with \t) is important and means that the element is a (html) child of the element above (with a lower index)
+- Elements tagged with `*[` are the new clickable elements that appeared on the website since the last step - if url has not changed.
+- Pure text elements without [] are not interactive.
+</browser_state>
+
+<browser_rules>
+Strictly follow these rules while using the browser and navigating the web:
+- Only interact with elements that have a numeric [index] assigned.
+- Only use indexes that are explicitly provided.
+- If the page changes after, for example, an input text action, analyze if you need to interact with new elements, e.g. selecting the right option from the list.
+- By default, only elements in the visible viewport are listed. Use scrolling actions if you suspect relevant content is offscreen which you need to interact with. Scroll ONLY if there are more pixels below or above the page.
+- You can scroll by a specific number of pages using the num_pages parameter (e.g., 0.5 for half page, 2.0 for two pages).
+- All the elements that are scrollable are marked with `data-scrollable` attribute. Including the scrollable distance in every directions. You can scroll *the element* in case some area are overflowed.
+- If a captcha appears, tell user you can not solve captcha. Finish the task and ask user to solve it.
+- If expected elements are missing, try scrolling, or navigating back.
+- If the page is not fully loaded, use the `wait` action.
+- Do not repeat one action for more than 3 times unless some conditions changed.
+- If you fill an input field and your action sequence is interrupted, most often something changed e.g. suggestions popped up under the field.
+- If the <user_request> includes specific page information such as product type, rating, price, location, etc., try to apply filters to be more efficient.
+- The <user_request> is the ultimate goal. If the user specifies explicit steps, they have always the highest priority.
+- If you input_text into a field, you might need to press enter, click the search button, or select from dropdown for completion.
+- Don't login into a page if you don't have to. Don't login if you don't have the credentials. 
+- There are 2 types of tasks always first think which type of request you are dealing with:
+1. Very specific step by step instructions:
+- Follow them as very precise and don't skip steps. Try to complete everything as requested.
+2. Open ended tasks. Plan yourself, be creative in achieving them.
+- If you get stuck e.g. with logins or captcha in open-ended tasks you can re-evaluate the task and try alternative ways, e.g. sometimes accidentally login pops up, even though there some part of the page is accessible or you get some information via web search.
+</browser_rules>
+
+<capability>
+- You can only handle single page app. Do not jump out of current page.
+- Do not click on link if it will open in a new page (e.g., <a target="_blank">)
+- It is ok to fail the task.
+	- User can be wrong. If the request of user is not achievable, inappropriate or you do not have enough information or tools to achieve it. Tell user to make a better request.
+	- Webpage can be broken. All webpages or apps have bugs. Some bug will make it hard for your job. It's encouraged to tell user the problem of current page. Your feedbacks (including failing) are valuable for user.
+	- Trying too hard can be harmful. Repeating some action back and forth or pushing for a complex procedure with little knowledge can cause unwanted results and harmful side-effects. User would rather you complete the task with a fail.
+- If you do not have knowledge for the current webpage or task. You must require user to give specific instructions and detailed steps.
+</capability>
+
+<task_completion_rules>
+You must call the `done` action in one of three cases:
+- When you have fully completed the USER REQUEST.
+- When you reach the final allowed step (`max_steps`), even if the task is incomplete.
+- When you feel stuck or unable to solve user request. Or user request is not clear or contains inappropriate content.
+- If it is ABSOLUTELY IMPOSSIBLE to continue.
+
+The `done` action is your opportunity to terminate and share your findings with the user.
+- Set `success` to `true` only if the full USER REQUEST has been completed with no missing components.
+- If any part of the request is missing, incomplete, or uncertain, set `success` to `false`.
+- You can use the `text` field of the `done` action to communicate your findings and to provide a coherent reply to the user and fulfill the USER REQUEST.
+- You are ONLY ALLOWED to call `done` as a single action. Don't call it together with other actions.
+- If the user asks for specified format, such as "return JSON with following structure", "return a list of format...", MAKE sure to use the right format in your answer.
+- If the user asks for a structured output, your `done` action's schema may be modified. Take this schema into account when solving the task!
+</task_completion_rules>
+
+<reasoning_rules>
+Exhibit the following reasoning patterns to successfully achieve the <user_request>:
+
+- Reason about <agent_history> to track progress and context toward <user_request>.
+- Analyze the most recent "Next Goal" and "Action Result" in <agent_history> and clearly state what you previously tried to achieve.
+- Analyze all relevant items in <agent_history> and <browser_state> to understand your state.
+- Explicitly judge success/failure/uncertainty of the last action. Never assume an action succeeded just because it appears to be executed in your last step in <agent_history>. If the expected change is missing, mark the last action as failed (or uncertain) and plan a recovery.
+- Analyze whether you are stuck, e.g. when you repeat the same actions multiple times without any progress. Then consider alternative approaches e.g. scrolling for more context or ask user for help.
+- Ask user for help if you have any difficulty. Keep user in the loop.
+- If you see information relevant to <user_request>, plan saving the information to memory.
+- Always reason about the <user_request>. Make sure to carefully analyze the specific steps and information required. E.g. specific filters, specific form fields, specific information to search. Make sure to always compare the current trajectory with the user request and think carefully if thats how the user requested it.
+</reasoning_rules>
+
+<examples>
+Here are examples of good output patterns. Use them as reference but never copy them directly.
+
+<evaluation_examples>
+"evaluation_previous_goal": "Successfully navigated to the product page and found the target information. Verdict: Success"
+"evaluation_previous_goal": "Clicked the login button and user authentication form appeared. Verdict: Success"
+</evaluation_examples>
+
+<memory_examples>
+"memory": "Found many pending reports that need to be analyzed in the main page. Successfully processed the first 2 reports on quarterly sales data and moving on to inventory analysis and customer feedback reports."
+</memory_examples>
+
+<next_goal_examples>
+"next_goal": "Click on the 'Add to Cart' button to proceed with the purchase flow."
+</next_goal_examples>
+</examples>
+
+<output>
+{
+  "evaluation_previous_goal": "Concise one-sentence analysis of your last action. Clearly state success, failure, or uncertain.",
+  "memory": "1-3 concise sentences of specific memory of this step and overall progress. You should put here everything that will help you track progress in future steps. Like counting pages visited, items found, etc.",
+  "next_goal": "State the next immediate goal and action to achieve it, in one clear sentence.",
+  "action":{
+    "Action name": {// Action parameters}
+  }
+}
+</output>

package/resources/local/page-agent/packages/core/src/tools/index.ts

@@ -0,0 +1,190 @@
+/**
+ * Internal tools for PageAgent.
+ * @note Adapted from browser-use
+ */
+import * as z from 'zod/v4'
+
+import type { PageAgentCore } from '../PageAgentCore'
+import { waitFor } from '../utils'
+
+/**
+ * Internal tool definition that has access to PageAgent `this` context
+ */
+export interface PageAgentTool<TParams = any> {
+	// name: string
+	description: string
+	inputSchema: z.ZodType<TParams>
+	execute: (this: PageAgentCore, args: TParams) => Promise<string>
+}
+
+export function tool<TParams>(options: PageAgentTool<TParams>): PageAgentTool<TParams> {
+	return options
+}
+
+/**
+ * Internal tools for PageAgent.
+ * Note: Using any to allow different parameter types for each tool
+ */
+export const tools = new Map<string, PageAgentTool>()
+
+tools.set(
+	'done',
+	tool({
+		description:
+			'Complete task. Text is your final response to the user — keep it concise unless the user explicitly asks for detail.',
+		inputSchema: z.object({
+			text: z.string(),
+			success: z.boolean().default(true),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			// @note main loop will handle this one
+			return Promise.resolve('Task completed')
+		},
+	})
+)
+
+tools.set(
+	'wait',
+	tool({
+		description: 'Wait for x seconds. Can be used to wait until the page or data is fully loaded.',
+		inputSchema: z.object({
+			seconds: z.number().min(1).max(10).default(1),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			// try to subtract LLM calling time from the actual wait time
+			const lastTimeUpdate = await this.pageController.getLastUpdateTime()
+			const actualWaitTime = Math.max(0, input.seconds - (Date.now() - lastTimeUpdate) / 1000)
+			console.log(`actualWaitTime: ${actualWaitTime} seconds`)
+			await waitFor(actualWaitTime)
+
+			return `✅ Waited for ${input.seconds} seconds.`
+		},
+	})
+)
+
+tools.set(
+	'ask_user',
+	tool({
+		description:
+			'Ask the user a question and wait for their answer. Use this if you need more information or clarification.',
+		inputSchema: z.object({
+			question: z.string(),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			if (!this.onAskUser) {
+				throw new Error('ask_user tool requires onAskUser callback to be set')
+			}
+			const answer = await this.onAskUser(input.question)
+			return `User answered: ${answer}`
+		},
+	})
+)
+
+tools.set(
+	'click_element_by_index',
+	tool({
+		description: 'Click element by index',
+		inputSchema: z.object({
+			index: z.int().min(0),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			const result = await this.pageController.clickElement(input.index)
+			return result.message
+		},
+	})
+)
+
+tools.set(
+	'input_text',
+	tool({
+		description: 'Click and type text into an interactive input element',
+		inputSchema: z.object({
+			index: z.int().min(0),
+			text: z.string(),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			const result = await this.pageController.inputText(input.index, input.text)
+			return result.message
+		},
+	})
+)
+
+tools.set(
+	'select_dropdown_option',
+	tool({
+		description:
+			'Select dropdown option for interactive element index by the text of the option you want to select',
+		inputSchema: z.object({
+			index: z.int().min(0),
+			text: z.string(),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			const result = await this.pageController.selectOption(input.index, input.text)
+			return result.message
+		},
+	})
+)
+
+/**
+ * @note Reference from browser-use
+ */
+tools.set(
+	'scroll',
+	tool({
+		description:
+			'Scroll vertically. Without index: scrolls the document. With index: scrolls the container at that index (or its nearest scrollable ancestor). Use index of a data-scrollable element to scroll a specific area.',
+		inputSchema: z.object({
+			down: z.boolean().default(true),
+			num_pages: z.number().min(0).max(10).optional().default(0.1),
+			pixels: z.number().int().min(0).optional(),
+			index: z.number().int().min(0).optional(),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			const result = await this.pageController.scroll({
+				...input,
+				numPages: input.num_pages,
+			})
+			return result.message
+		},
+	})
+)
+
+/**
+ * @todo Tables need a dedicated parser to extract structured data. This tool is useless.
+ */
+tools.set(
+	'scroll_horizontally',
+	tool({
+		description:
+			'Scroll horizontally. Without index: scrolls the document. With index: scrolls the container at that index (or its nearest scrollable ancestor). Use index of a data-scrollable element to scroll a specific area.',
+		inputSchema: z.object({
+			right: z.boolean().default(true),
+			pixels: z.number().int().min(0),
+			index: z.number().int().min(0).optional(),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			const result = await this.pageController.scrollHorizontally(input)
+			return result.message
+		},
+	})
+)
+
+tools.set(
+	'execute_javascript',
+	tool({
+		description:
+			'Execute JavaScript code on the current page. Supports async/await syntax. Use with caution!',
+		inputSchema: z.object({
+			script: z.string(),
+		}),
+		execute: async function (this: PageAgentCore, input) {
+			const result = await this.pageController.executeJavascript(input.script)
+			return result.message
+		},
+	})
+)
+
+// @todo send_keys
+// @todo upload_file
+// @todo go_back
+// @todo extract_structured_data

package/resources/local/page-agent/packages/core/src/types.ts

@@ -0,0 +1,289 @@
+import type { LLMConfig } from '@page-agent/llms'
+
+// @note circular dependency but okay
+import type { PageAgentCore } from './PageAgentCore'
+import type { PageAgentTool } from './tools'
+
+/** Supported UI languages */
+export type SupportedLanguage = 'en-US' | 'zh-CN'
+
+export interface AgentConfig extends LLMConfig {
+	language?: SupportedLanguage
+
+	/**
+	 * Maximum number of steps the agent can take per task.
+	 * @default 40
+	 */
+	maxSteps?: number
+
+	/**
+	 * Custom tools to extend PageAgent capabilities
+	 * @experimental
+	 * @note You can also override or remove internal tools by using the same name.
+	 * @see PageAgentTool
+	 *
+	 * @example
+	 * // override internal tool
+	 * import { z } from 'zod/v4'
+	 * import { tool } from 'page-agent'
+	 * const customTools = {
+	 * ask_user: tool({
+	 * 	description:
+	 * 		'Ask the user or parent model a question and wait for their answer. Use this if you need more information or clarification.',
+	 * 	inputSchema: z.object({
+	 * 		question: z.string(),
+	 * 	}),
+	 * 	execute: async function (this: PageAgent, input) {
+	 * 		const answer = await do_some_thing(input.question)
+	 * 		return "✅ Received user answer: " + answer
+	 * 	},
+	 * })
+	 * }
+	 *
+	 * @example
+	 * // remove internal tool
+	 * const customTools = {
+	 * 	ask_user: null // never ask user questions
+	 * }
+	 */
+	customTools?: Record<string, PageAgentTool | null>
+
+	/**
+	 * Instructions to guide the agent's behavior
+	 */
+	instructions?: {
+		/**
+		 * Global system-level instructions, applied to all tasks
+		 */
+		system?: string
+
+		/**
+		 * Dynamic page-level instructions callback
+		 * Called before each step to get instructions for the current page
+		 * @param url - Current page URL (window.location.href)
+		 * @returns Instructions string, or undefined/null to skip
+		 */
+		getPageInstructions?: (url: string) => string | undefined | null
+	}
+
+	/**
+	 * Lifecycle hooks for task execution.
+	 * @experimental API may change in future versions.
+	 *
+	 * All hooks receive the agent instance as first parameter.
+	 */
+
+	/**
+	 * Called before each step execution.
+	 * @experimental
+	 * @param agent - The PageAgentCore instance
+	 * @param stepCount - Current step number (0-indexed)
+	 */
+	onBeforeStep?: (agent: PageAgentCore, stepCount: number) => Promise<void> | void
+
+	/**
+	 * Called after each step execution.
+	 * @experimental
+	 * @param agent - The PageAgentCore instance
+	 * @param history - Current history of events
+	 */
+	onAfterStep?: (agent: PageAgentCore, history: HistoricalEvent[]) => Promise<void> | void
+
+	/**
+	 * Called before task execution starts.
+	 * @experimental
+	 * @param agent - The PageAgentCore instance
+	 */
+	onBeforeTask?: (agent: PageAgentCore) => Promise<void> | void
+
+	/**
+	 * Called after task execution completes (success or failure).
+	 * @experimental
+	 * @param agent - The PageAgentCore instance
+	 * @param result - The execution result
+	 */
+	onAfterTask?: (agent: PageAgentCore, result: ExecutionResult) => Promise<void> | void
+
+	/**
+	 * Called when the agent is disposed.
+	 * @experimental
+	 * @note This hook can block the disposal process if it's async.
+	 * @param agent - The PageAgentCore instance
+	 * @param reason - Optional reason for disposal
+	 */
+	onDispose?: (agent: PageAgentCore, reason?: string) => void
+
+	// page behavior hooks
+
+	/**
+	 * @experimental
+	 * Enable the experimental script execution tool that allows executing generated JavaScript code on the page.
+	 * @note Can cause unpredictable side effects.
+	 * @note May bypass some safe guards and data-masking mechanisms.
+	 */
+	experimentalScriptExecutionTool?: boolean
+
+	/**
+	 * @experimental
+	 * Fetch /llms.txt from current site origin and include as context.
+	 * Only fetched once per origin per task.
+	 * @default false
+	 */
+	experimentalLlmsTxt?: boolean
+
+	/**
+	 * Transform page content before sending to LLM.
+	 * Called after DOM extraction and simplification, before LLM invocation.
+	 * Use cases: inspect extraction results, modify page info, mask sensitive data.
+	 *
+	 * @param content - Simplified page content that will be sent to LLM
+	 * @returns Transformed content
+	 *
+	 * @example
+	 * // Mask phone numbers
+	 * transformPageContent: async (content) => {
+	 *   return content.replace(/1[3-9]\d{9}/g, '***********')
+	 * }
+	 */
+	transformPageContent?: (content: string) => Promise<string> | string
+
+	/**
+	 * Completely override the default system prompt.
+	 * @experimental Use with caution - incorrect prompts may break agent behavior.
+	 */
+	customSystemPrompt?: string
+
+	/**
+	 * Delay between steps in seconds.
+	 * @default 0.4
+	 */
+	stepDelay?: number
+}
+
+/**
+ * Agent reflection state - the reflection-before-action model
+ *
+ * Every tool call must first reflect on:
+ * - evaluation_previous_goal: How well did the previous action achieve its goal?
+ * - memory: Key information to remember for future steps
+ * - next_goal: What should be accomplished in the next action?
+ */
+export interface AgentReflection {
+	evaluation_previous_goal: string
+	memory: string
+	next_goal: string
+}
+
+/**
+ * MacroTool input structure
+ *
+ * This is the core abstraction that enforces the "reflection-before-action" mental model.
+ * Before executing any action, the LLM must output its reasoning state.
+ */
+export interface MacroToolInput extends Partial<AgentReflection> {
+	action: Record<string, any>
+}
+
+/**
+ * MacroTool output structure
+ */
+export interface MacroToolResult {
+	input: MacroToolInput
+	output: string
+}
+
+/**
+ * A single agent step with reflection and action
+ */
+export interface AgentStepEvent {
+	type: 'step'
+	stepIndex: number
+	reflection: Partial<AgentReflection>
+	action: {
+		name: string
+		input: any
+		output: string
+	}
+	usage: {
+		promptTokens: number
+		completionTokens: number
+		totalTokens: number
+		cachedTokens?: number
+		reasoningTokens?: number
+	}
+	/** Raw LLM response for debugging */
+	rawResponse?: unknown
+	/** Raw LLM request for debugging */
+	rawRequest?: unknown
+}
+
+/**
+ * Persistent observation event (stays in memory)
+ */
+export interface ObservationEvent {
+	type: 'observation'
+	content: string
+}
+
+/**
+ * User takeover event
+ */
+export interface UserTakeoverEvent {
+	type: 'user_takeover'
+}
+
+/**
+ * Retry event - LLM call is being retried
+ */
+export interface RetryEvent {
+	type: 'retry'
+	message: string
+	attempt: number
+	maxAttempts: number
+}
+
+/**
+ * Error event - fatal error from LLM or execution
+ */
+export interface AgentErrorEvent {
+	type: 'error'
+	message: string
+	rawResponse?: unknown
+}
+
+/**
+ * Union type for all history events
+ */
+export type HistoricalEvent =
+	| AgentStepEvent
+	| ObservationEvent
+	| UserTakeoverEvent
+	| RetryEvent
+	| AgentErrorEvent
+
+/**
+ * Agent execution status
+ */
+export type AgentStatus = 'idle' | 'running' | 'completed' | 'error'
+
+/**
+ * Agent activity - transient state for immediate UI feedback.
+ *
+ * Unlike historical events (which are persisted), activities are ephemeral
+ * and represent "what the agent is doing right now". UI components should
+ * listen to 'activity' events to show real-time feedback.
+ *
+ * Note: There is no 'idle' activity - absence of activity events means idle.
+ */
+export type AgentActivity =
+	| { type: 'thinking' }
+	| { type: 'executing'; tool: string; input: unknown }
+	| { type: 'executed'; tool: string; input: unknown; output: string; duration: number }
+	| { type: 'retrying'; attempt: number; maxAttempts: number }
+	| { type: 'error'; message: string }
+
+export interface ExecutionResult {
+	success: boolean
+	data: string
+	history: HistoricalEvent[]
+}