菜单
Use delegate_task with multiple browser-equipped sub-agents to extract verbatim text from Chinese academic/religious websites with JavaScript-heavy tree navigation. Covers coordination patterns, site-specific navigation, error recovery, and QA verification.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when Hermes Agent needs X/Twitter search, public signal research, monitoring, or approval-gated account actions through the Hermes Tweet plugin.
Router Learning System — 越用越聰明嘅任務路由引擎。 每個 task 有最佳 path，Router 自動記錄經驗、動態調整、跳過爛 path。
Use when signaling work status across autonomous sessions.
Use when needing to standardize descriptions across many skill files at once — change format, fix wording, or enforce a convention like 'Use when...' prefix.
Use before making irreversible changes (deleting files, destructive git operations, large refactors). Use when you need the ability to roll back if something goes wrong. Use at the start of any task that modifies multiple files.
Use when the context window is getting full and tasks need breaking down.
| name | multi-agent-browser-text-extraction |
| description | Use delegate_task with multiple browser-equipped sub-agents to extract verbatim text from Chinese academic/religious websites with JavaScript-heavy tree navigation. Covers coordination patterns, site-specific navigation, error recovery, and QA verification. |
| metadata | {"hermes":{"platform":["cursor","codex","claude-code"]},"triggers":["browser extract","scrape","web content","extract page"]} |
| domain | ecosystem |
| subdomain | agent-skills |
| tokens | {"scan":195,"load":1680,"category":"detailed"} |
Extract large volumes of original text from JavaScript-heavy Chinese websites (e.g. ddc.shengyen.org 法鼓全集) using parallel delegate_task sub-agents, each with browser toolset.
delegate_task tool with browser toolset enabledhttps://ddc.shengyen.org/)javascript:void(null) links (they usually are)browser_console to extract content:
document.querySelector('[class*="content"]')?.innerText || document.body.innerText.substring(0, 5000)
Divide the chapters into batches of 5-9 questions each. Each batch spawns a delegate_task with toolsets: ["browser"].
Critical passing of context:
05-02 正信的佛教)max_iterations=40-55 for 6-9 chaptersContext template:
HOW TO USE THE SITE:
1. Go to https://ddc.shengyen.org/
2. Click "第五輯 佛教入門類" in left sidebar to expand
3. Click the book (e.g. "05-03 學佛群疑") to expand its chapters
4. Click a chapter name from the expanded sub-menu
5. Use browser_console: document.querySelector('[class*="content"]')?.innerText || ...
6. Copy the EXACT text, verbatim
QUESTIONS TO EXTRACT:
{list of questions with book and chapter mapping}
Output format for EACH question:
## Question: {question}
**Source**: 《書名》聖嚴法師 — Chapter Name
**Original Text**:
> {exact, verbatim, can be long}
After all sub-agents complete:
javascript:void(null) links — browser snapshot won't show chapter content directly. You MUST use browser_console with innerText extraction after clicking.browser_vision to find which element index to click.ddm.org.tw blocks automated browsers. The mirror ddc.shengyen.org works without blocking.max_iterations generously (40-55) for 6-9 chapters. If still failing, split into smaller batches.context parameter. Don't assume they know project structure.After compiling, run a quick check:
const texts = require('./data/quiz-original-texts.js');
const targetIds = [...]; // all expected IDs
targetIds.forEach(id => {
if (!texts[id] || texts[id].length === 0) console.error(`MISSING: ${id}`);
});
Extracting 21 chapters from Master Shengyan's 法鼓全集 at ddc.shengyen.org: