agent-browser

agent-browser Business Diagnostic Assistant

Quick Start

More Skills

Overview

The business diagnostic assistant is used to answer "how should this be done," conduct web research, public information research, and competitor analysis, and is suitable for marketing, operations, product, and sales to invoke after clarifying business goals, content materials, or analysis targets. It combines inputs such as browsing research context, pasted research task descriptions, visible text from public pages, candidate links, step records, screenshot transcriptions, or... to organize key context, and outputs summaries, diagnostic conclusions, action recommendations, and reusable deliverables for continued execution, review, or delivery.

When to Use

Applicable Scenarios

• Users ask questions such as "how should this be done," need to quickly break down goals, determine priorities, and form executable results
• Marketing, operations, product, and sales need to generate summaries, diagnostic conclusions, action recommendations, and reusable deliverables around the business diagnostic assistant
• Users have already prepared agentic browsing goals (for example, data collection, page comparison, process observation, evidence archiving, or manual execution playbooks.), runbook readers (fill in users such as researchers, QA, operations, product managers, sales intelligence teams, content teams, etc.), public entry links (fill in entry URLs accessible without login, candidate page links, page titles, or link metadata.), and hope to organize them into executable analysis or optimization results
• Users need to turn materials related to the business diagnostic assistant into clear conclusions, priorities, and next steps

How to Invoke

Call the AI Skills API directly through the exported Python runner:

Command Example

Basic Invocation

python3 scripts/run.py --params '{}'

Invocation with Common Parameters

python3 scripts/run.py --params '{"goal":"代理式浏览目标"}'

Parameter Description

Parameter	Type	Required	Default	Description
goal	string	No	-	For example, data collection, page comparison, process observation, evidence archiving, or manual execution playbooks
audience	string	No	-	Fill in users such as researchers, QA, operations, product managers, sales intelligence teams, content teams, etc.
materialUrl	string	No	-	Fill in entry URLs accessible without login, candidate page links, page titles, or link metadata; a complete accessible URL must be provided
reviewDepth	string	No	标准诊断	Choose a quick task card, standard runbook, or deep research protocol; optional values:快速体检、标准诊断、深度改稿
materialFile	string	No	-	Upload task materials, page observation records, research packages, PDF, Markdown, CSV, TXT, or screenshot transcriptions
materialText	string	No	-	Paste research task descriptions, visible text from public pages, candidate links, step records, screenshot transcriptions, or observation notes
targetPlatform	string	No	-	Explain whether the plan is intended for official websites, help centers, public directories, documentation sites, public SaaS pages, or other page types
brandRequirements	string	No	-	Supplement brand voice, compliance boundaries, prohibited operations, citation standards, privacy requirements, or output format

See the complete machine-readable parameter structure at references/form-schema.json。

Parameter value reference

This skill currently has no categorical parameters that require looking up additional tables.

Supported input formats

This skill directly accepts JSON parameters; if the parameters include a link field, please provide a complete, accessible URL.

Example request

The example parameters below can be passed directly to scripts/run.py, and the runner will send them to the AI Skills API.

python3 scripts/run.py --params '{"goal":"代理式浏览目标"}'

Equivalent --params JSON：

{
  "goal": "代理式浏览目标"
}

Example return result

{
  "success": true,
  "data": {
    "message": "示例结果请以技能真实返回结构为准。"
  },
  "meta": {
    "executionTime": 842,
    "cached": false
  }
}

Deliverables

• Summary, diagnostic conclusions, action recommendations, and reusable deliverables: organize the core results around user goals that can be directly read, reviewed, or delivered.
• Interpretation of input materials: combine the agentic browsing goal (for example, information collection, page comparison, process observation, evidence archiving, or a manual execution handbook.), the runbook reader (fill in users such as researchers, QA, operations, product managers, sales intelligence teams, content teams, etc.), and public entry links (fill in entry URLs accessible without login, candidate page links, page titles, or link metadata.) to extract key context and the basis for judgment.
• Next steps: provide priorities, execution recommendations, or a content framework that can be further processed.

Suggestions for using the results

• First determine whether the output answers the user's core questions about the "Business Diagnostic Assistant".
• Then check whether the results cover the summary, diagnostic conclusions, action recommendations, and reusable deliverables, and whether clear next steps are provided.
• If the input materials are limited, it is recommended to have the user supplement the goals, audience, constraints, or original materials before running.

Preparation before running

• AISKILLS_BASE_URL: default https://ai-skills.ai
• AISKILLS_API_KEY: required, used for authenticated calls
• AISKILLS_TENANT_ID: default default