AGENT SKILL

microsoft--webwright

Browser agent mã nguồn mở của Microsoft Research — agent viết Python/Playwright script thay vì click từng bước, script có thể reuse. SOTA 86.7% trên Online-Mind2Web.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 13
Forks
2

01 前言 Front Matter

name
microsoft--webwright
description
Browser agent mã nguồn mở của Microsoft Research — agent viết Python/Playwright script thay vì click từng bước, script có thể reuse. SOTA 86.7% trên Online-Mind2Web.
allowed-tools
Bash, Read, Write
user-invocable
true

03 SKILL.md 正文

Webwright là browser agent framework: thay vì predict discrete actions, agent viết Python/Playwright script tự do — script reusable, testable, maintainable như code thật.

Install

pip install -e .
playwright install chromium

Usage

python -m webwright.run.cli \
    -c base.yaml -c model_openai.yaml \
    -t "Book a flight from Hanoi to HCM City, cheapest option" \
    --start-url https://www.google.com/travel/flights \
    -o outputs/flight-booking

Config

# base.yaml
workspace_dir: ./workspace
max_steps: 50
screenshot_on_failure: true

# model_anthropic.yaml
model:
  provider: anthropic
  name: claude-sonnet-4-6
  api_key: ${ANTHROPIC_API_KEY}

Core Pattern — Agent Loop

Task → Launch browser → Observe (screenshot + DOM)
     → Write Playwright script → Execute → Inspect failure?
     → Screenshot + re-observe → Refine script → Loop
     → Final artifact = reusable Python script + logs

Slash Commands (Claude Code plugin)

/plugin install webwright@webwright

/webwright:run    -- single-shot task
/webwright:craft  -- tạo reusable CLI tool từ task description

Điểm mạnh

  • Script output là code thật — không phải click trace
  • 86.7% Online-Mind2Web SOTA (2026)
  • ~1.5k lines core — đọc hiểu dễ
  • Support OpenAI, Anthropic, OpenRouter
  • Trajectory + screenshots lưu disk để debug

Khi nào dùng

  • Automation web phức tạp (multi-step, conditional logic)
  • Cần script có thể chạy lại, không phải one-off recording
  • RPA-style workflow với LLM decision layer
  • Web scraping có login, dynamic content

Source

https://github.com/microsoft/Webwright · MIT

04 同仓库更多 Skills

Horonomos-laps
Horonomos-laps/Yana-AI

openai--sentry--sentry

Use when the user asks to inspect Sentry issues or events, summarize recent production errors, or pull basic Sentry health data via the Sentry API; perform read-only queries with the bundled script and require `SENTRY_AUTH_TOKEN`.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

ito-basket-compare

Compare Itô prediction-market baskets against a user's knowledge base, portfolio notes, financial context, watchlist, or research thesis. Use for read-only basket comparison and gap analysis without investment advice or live trading.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

implementing-network-traffic-baselining

Build network traffic baselines from NetFlow/IPFIX data using Python pandas for statistical analysis, z-score anomaly detection, and hourly/daily traffic pattern profiling

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

detecting-evasion-techniques-in-endpoint-logs

Detects defense evasion techniques used by adversaries in endpoint logs including log tampering, timestomping, process injection, and security tool disabling. Use when investigating suspicious endpoint behavior, building detection rules for evasion tactics, or conducting threat hunting for stealthy adversary activity. Activates for requests involving evasion detection, defense evasion analysis, log tampering detection, or MITRE ATT&CK TA0005.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

implementing-microsegmentation-with-guardicore

Implementing microsegmentation using Akamai Guardicore Segmentation to map application dependencies, create granular network policies, visualize east-west traffic flows, and enforce least-privilege communication between workloads across data centers and cloud.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

ramdisk-tmpfs-patterns

暂无简介

★ 13