Skills Plugins MCP Prompt Model 博客 我的中心
生活とツール #browser #automation #ai #agent

stealth-browser-mcp

Use this skill when operating stealth-browser-mcp from an AI agent or MCP client for browser automation, page inspection, element interaction, screenshots, file uploads, CDP commands, network debugging, cookies/storage, stealth setup, or reliable multi-step browser workflows. It provides the correct tool order, state checks, lifecycle cleanup, and pre-document script guidance.

DeepseekModel キュレーション済みスキル 品質 優秀 · 90 v1.0.0

取得

https://deepseekmodel.com/api/download.php?id=vibheksoni-stealth-browser-mcp-skills-stealth-browser-mcp-skill-md&format=skill
ダウンロード .skill 標準形式。system_prompt と model_config を収録し、任意の Agent で利用可能
.skill ファイルの system_prompt フィールドの実際の内容。
name stealth-browser-mcp description Use this skill when operating stealth-browser-mcp from an AI agent or MCP client for browser automation, page inspection, element interaction, screenshots, file uploads, CDP commands, network debugging, cookies/storage, stealth setup, or reliable multi-step browser workflows. It provides the correct tool order, state checks, lifecycle cleanup, and pre-document script guidance. Stealth Browser MCP Core Workflow Start with spawn_browser and keep the returned instance_id for every follow-up tool call. Use navigate for page loads, then verify state with get_instance_state , get_page_content , or take_screenshot . Inspect before acting. Use query_elements , get_page_content , and screenshots to find stable selectors and confirm the target element is visible. Act with the narrowest tool that fits the task, such as click_element , type_text , scroll_page , wait_for_element , or file_upload . Verify after each important action with page content, DOM queries, URL/title state, or a screenshot. Close browsers with close_instance when the task is complete unless the user explicitly wants the session left open. Tool Selection Use --minimal when an agent only needs browser spawning, navigation, screenshots, page content, and basic interaction. Use full mode for CDP execution, network debugging, cookie/storage work, dynamic hooks, and advanced element extraction. Use execute_script for normal page JavaScript. Use execute_cdp_command only when a CDP primitive is required or when page JavaScript cannot access the needed browser state. Use network tools after navigation starts: list_network_requests , search_network_requests , get_request_details , get_response_details , and get_response_content . Use extraction tools when the task is to recreate, clone, or analyze page elements with styles, assets, events, or animations. Pre-Document Scripts Use add_script_to_evaluate_on_new_document before navigation when a site must see modified browser APIs during its first inline scripts. This is required for WebGL renderer/vendor spoofing, navigator property patches, and similar anti-bot checks that run during page parse. Keep world_name=None for main-world browser API spoofing. Set run_immediately=True when the patch should also run in the current document. Reliability Rules Prefer stable CSS selectors, roles, visible text, or attributes over brittle generated classes. Wait for page state instead of assuming navigation or rendering has finished. Treat every tool result as stateful: pass the exact instance_id , request IDs, and selectors returned by prior steps. For file uploads, pass absolute file paths inside directories allowed by BROWSER_FILE_UPLOAD_ALLOWED_DIRS . If output is too large, ask for a narrower selector, use filters, or rely on file-backed extraction responses. Keep debug logging disabled during normal MCP stdio usage. Enable --debug or STEALTH_BROWSER_DEBUG=1 only when diagnosing server behavior. Lifecycle And Security The server reaps idle browsers by default. BROWSER_IDLE_TIMEOUT=0 disables idle reaping globally. HTTP transport is unauthenticated by default for compatibility. Set STEALTH_BROWSER_MCP_AUTH_TOKEN when exposing HTTP transport beyond a trusted local environment. Do not expose unauthenticated HTTP transport to public networks. Close instances after tests and one-off automations to release browser processes and temporary profiles.
このスキルを起動するキーワード。クリックでコピーできます。

このスキルにはトリガーワードがありません。

ダウンロードした .skill に含まれるフィールド。
フィールド 説明
formatフォーマット識別子(skill/v1)
skill_idスキル固有 ID
nameスキル名
versionバージョン
description説明
categoryカテゴリ(配列)
trigger_wordsトリガーワード
tagsタグ
sourceソース
source_urlソース URL(本ページ)
exported_atエクスポート日時(ダウンロード毎)
system_promptシステムプロンプト本文
model_configモデル設定:provider / model / temperature / max_tokens / top_p
examplesサンプル
install_guide各プラットフォームの導入説明(Coze / Dify / Claude / カスタム)
同じスキルを各プラットフォーム形式で出力できます。
.skill 標準形式。system_prompt と model_config を収録し、任意の Agent で利用可能 ダウンロード
.skillpro 拡張形式。scripts / tools / dependencies / hooks を含む ダウンロード
.json 純粋な JSON 出力。system_prompt とモデル設定のみ ダウンロード
Coze frontmatter 付き Markdown。Coze へのインポート用 ダウンロード
Dify Dify DSL。アプリ作成後にそのままインポート ダウンロード

每日精选 Skill 推荐,免费送到你邮箱

输入邮箱,每天接收一个精选 AI Agent 技能推荐。完全免费,持续更新。

提交后我们会发送一封确认邮件,点击邮件里的链接才会开始收信。

完全免费,取消任意时间。我们不会发送垃圾邮件。