下载地址
https://www.zhengdianxue.com/wp-content/uploads/skills/indiv-sayelf--agent-browser-core.zip将提示词发送给你的 AI 安装该 skills
请根据 https://skillhub.cn/install/skillhub.md,安装 @indiv-sayelf/agent-browser-core。OpenClaw skill for the agent-browser CLI (Rust-based with Node.js fallback) enabling AI-friendly web automation with snapshots, refs, and structured commands.
阅读全文 (展开查看完整说明)
概述
Agent Browser Skill (Core)
Purpose
Provide an advanced, production-ready playbook for using agent-browser to automate web tasks via CLI and structured commands.
Quick orientation
- Read
references/agent-browser-overview.mdfor install, architecture, and core concepts. - Read
references/agent-browser-command-map.mdfor command categories and flags. - Read
references/agent-browser-safety.mdfor high-risk controls and safe mode rules. - Read
references/agent-browser-workflows.mdfor recommended AI workflows. - Read
references/agent-browser-troubleshooting.mdfor common issues and fixes.
Required inputs
- Installed agent-browser CLI and browser runtime.
- Target URLs and workflow steps.
- Session or profile strategy if authentication is required.
Expected output
- A clear command sequence and operational guardrails for automation.
Operational notes
- Snapshot early, act via refs, then snapshot again after DOM changes.
- Use
--jsonfor machine parsing and scripting. - Use waits and load-state checks before actions.
- Close tabs or sessions when done to release resources.
Safe mode defaults
- Do not use
eval,--allow-file-access, custom--executable-path, or arbitrary--argswithout explicit approval. - Avoid
network route,set credentials, and cookie/storage mutations unless the task requires it. - Allowlist domains and block localhost or private network targets.
Security notes
- Treat tokens and credentials as secrets.
- Avoid
--allow-file-accessunless explicitly required.
Session & auth playbook
- Session Isolation: By default,
agent-browserlaunches an independent browser instance and does NOT share the user's existing login sessions. - Reusing Profiles: Use
--user-data-dirto persist and reuse login states across runs. - Connecting to Existing Chrome: Use
--cdpto attach to an already running Chrome instance (useful for bypassing complex auth). - Auth Fallback: For complex OAuth or Cloudflare CAPTCHA scenarios that block automation, immediately hand-off to the user for manual intervention rather than infinitely retrying.
Complex task orchestration
- Step-by-Step Execution: For complex tasks (e.g., batch downloads, multi-page scraping, complex forms), first output a clear plan, then execute step-by-step, reporting the result of each step before proceeding.
- Pacing and Human-likeness: Control operation pacing. Introduce random delays or simulate human input speed to avoid triggering anti-bot mechanisms.
- Sensitive Operations: Explicitly ask for user confirmation before executing sensitive actions such as entering passwords, initiating batch downloads, or deleting data.
Result verification
- Data Validation: If scraped or extracted fields return empty, zero (
0.00), or abnormal values, proactively prompt the user to confirm the data source or page state. - Write Operation Re-check: After executing any write operation (click, fill, submit), you MUST re-snapshot or use
getto verify the DOM state has actually changed. - No Fake Completions: Strictly prohibited to output "Task Completed" or "Updated Successfully" if the underlying data or page state has not substantively changed.