reverse-skills-max: initial public release (restored, redacted, hardened)

This commit is contained in:
baradok609hecl1pq6qi
2026-10-20 02:17:15 +03:00
commit da24c2bd20
2227 changed files with 443931 additions and 0 deletions

View File

@@ -0,0 +1,224 @@
#!/usr/bin/env node
// 看雪论坛 / 微信公众号技术文章源获取工具(方案1:直接抓取官方论坛源)
//
// 背景:看雪学苑公众号文章均为看雪论坛(bbs.kanxue.com)帖子转载,微信侧无官方
// API 可枚举他人公众号文章,因此直接抓论坛列表即等价于公众号选文源。
//
// 子命令:
// kanxue list [--board 名|--id N] [--pages N] [--md] 论坛文章列表(默认全站)
// kanxue thread <帖子ID> [--md] 单篇帖子标题/作者/正文
// wechat <文章URL> [--md] 微信文章抓取(备用渠道)
//
// 输出默认 JSON(结构化,便于管道/喂给 AI);--md 输出 markdown(便于直接入库)。
// 零依赖(Node >=22 内置 fetch)。导出纯函数供 tests/wxsource.test.mjs 测试。
const UA =
'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0 Safari/537.36';
// 版块名(中文/英文别名)→ forum id(首页导航实测确认;未知版块用 --id 指定)
const BOARDS = {
home: null, // 全站首页
ai: 170, 'ai-help': 170, 'ai助力安全': 170, 'ai 助力安全': 170,
'ai-self': 168, 'ai自身安全': 168, 'ai 自身安全': 168,
'ai-platform': 123, 'ai平台与数据安全': 123,
re: 4, 逆向工程: 4, 软件逆向: 4,
ctf: 37, 'ctf对抗': 37,
crypto: 132, 密码应用: 132, 密码算法: 132,
tools: 10, 安全工具: 10,
harmony: 178, 鸿蒙: 178,
android: 161, 'android安全': 161,
iot: 128, 'iot安全': 128,
ios: 166, 'ios安全': 166,
coding: 41, 编程技术: 41,
unpack: 88, 加壳脱壳: 88,
vuln: 150, 二进制漏洞: 150,
pwn: 171,
web: 151, 'web安全': 151,
translate: 32, 外文翻译: 32,
};
export const BASE = 'https://bbs.kanxue.com';
// ---------- 网络 ----------
export async function fetchText(url, { retries = 1 } = {}) {
for (let i = 0; ; i++) {
const ctrl = new AbortController();
const timer = setTimeout(() => ctrl.abort(), 20000);
try {
const res = await fetch(url, {
headers: { 'user-agent': UA, 'accept-language': 'zh-CN,zh;q=0.9' },
signal: ctrl.signal,
});
if (!res.ok) throw new Error(`HTTP ${res.status} ${url}`);
return await res.text();
} catch (e) {
if (i >= retries) throw e;
} finally {
clearTimeout(timer);
}
}
}
// ---------- HTML → 简化 markdown 文本 ----------
export function htmlToText(html) {
let s = html
.replace(/<script[\s\S]*?<\/script>/gi, '')
.replace(/<style[\s\S]*?<\/style>/gi, '')
.replace(/<br\s*\/?>/gi, '\n')
.replace(/<\/(p|div|section|li|h[1-6]|pre|tr|blockquote)>/gi, '\n')
.replace(/<img[^>]*>/gi, '[图片]')
.replace(/<code[^>]*>/gi, '`')
.replace(/<\/code>/gi, '`')
.replace(/<li[^>]*>/gi, '- ')
.replace(/<[^>]+>/g, '');
s = s
.replace(/&nbsp;/g, ' ')
.replace(/&lt;/g, '<')
.replace(/&gt;/g, '>')
.replace(/&amp;/g, '&')
.replace(/&quot;/g, '"')
.replace(/&#39;/g, "'");
return s.replace(/\n{3,}/g, '\n\n').replace(/[ \t]+\n/g, '\n').trim();
}
// ---------- 看雪论坛列表解析 ----------
// 首页聚合列表:<a class="bbs_home_page_list_title" href="thread-292402.htm" title="...">标题</a>
export function parseKanxueHomeList(html) {
const items = [];
const re =
/<a[^>]*class="[^"]*bbs_home_page_list_title[^"]*"[^>]*href="(thread-\d+\.htm)"[^>]*title="([^"]*)"[^>]*>[\s\S]*?<\/a>/g;
for (const m of html.matchAll(re)) {
items.push({ url: `${BASE}/${m[1]}`, title: m[2].trim() || htmlToText(m[0]) });
}
return items;
}
// 分类页(xiuno):<tr class="thread ..." data-tid="292346"> 行内标题 + 作者
export function parseForumList(html) {
const items = [];
const re = /<tr[^>]*class="thread[^"]*"[^>]*data-tid="(\d+)"[\s\S]*?<\/tr>/g;
for (const m of html.matchAll(re)) {
const tid = m[1];
const t = m[0].match(/<a href="thread-\d+\.htm"[^>]*>([^<]{4,120})<\/a>/);
const au = [...m[0].matchAll(/<a href="user-home-\d+\.htm"[^>]*>([^<]{2,40})<\/a>/g)].at(-1);
if (!t) continue;
items.push({
url: `${BASE}/thread-${tid}.htm`,
title: htmlToText(t[1]),
author: au ? htmlToText(au[1]) : undefined,
});
}
return items;
}
// ---------- 看雪帖子正文解析 ----------
export function parseThread(html) {
const title = html.match(/<title>([^<]*)</)?.[1]?.replace(/-[^-]*-看雪安全社区.*$/, '').trim();
const author = html.match(/<a href="user-home-\d+\.htm"[^>]*title="([^"]+)"[^>]*>/)?.[1] ||
html.match(/user-home-\d+\.htm"[^>]*>([^<]{2,40})<\/a>/)?.[1];
// 正文:优先取第一条消息(isfirst=1,markdown 或富文本类型均可),终止于下一条消息
const first =
html.match(/<div class="message[^"]*"[^>]*isfirst="1"[^>]*>([\s\S]*?)<\/div>\s*<div class="message/) ||
html.match(/<div class="message[^"]*"[^>]*>([\s\S]*?)<\/div>\s*<div class="message/) ||
html.match(/<div class="message[^"]*"[^>]*isfirst="1"[^>]*>([\s\S]*)/) ||
html.match(/<div class="message[^"]*"[^>]*>([\s\S]*)/);
return { title: title || undefined, author, body: htmlToText(first ? first[1] : '') };
}
// ---------- 微信文章解析(备用渠道) ----------
export function parseWechat(html) {
const meta = (prop) => html.match(new RegExp(`<meta property="${prop}" content="([^"]*)"`))?.[1];
const title = meta('og:title');
const author = meta('og:article:author') || meta('og:description');
const body = html.match(/<div[^>]*id="js_content"[^>]*>([\s\S]*?)(?:<\/div>\s*<script|$)/)?.[1] || '';
return { title, author, body: htmlToText(body) };
}
// ---------- CLI ----------
function usage() {
console.log(`看雪/微信技术文章源获取工具
用法:
wxsource.mjs kanxue list [--board 名|--id N] [--pages N] [--md]
论坛文章列表(默认全站首页;--md 输出 markdown 列表)
内置版块: ${Object.keys(BOARDS).filter((k) => BOARDS[k]).join(' / ')}
wxsource.mjs kanxue thread <帖子ID> [--md]
单篇帖子标题/作者/正文(帖子ID 如 292402)
wxsource.mjs wechat <文章URL> [--md]
微信文章抓取(备用渠道,公众号内容 = 论坛转载)
输出默认 JSON({url,title,author,body} 结构),--md 输出 markdown。`);
}
function argVal(args, name, fallback) {
const i = args.indexOf(name);
return i >= 0 && args[i + 1] !== undefined ? args[i + 1] : fallback;
}
function has(args, name) {
return args.includes(name);
}
async function main(argv) {
const args = argv.slice(2);
if (args.length === 0 || has(args, '-h') || has(args, '--help')) return usage();
const md = has(args, '--md');
const cmd = args[0];
if (cmd === 'kanxue' && args[1] === 'list') {
const board = argVal(args, '--board', 'home');
const boardId = argVal(args, '--id', null) ?? BOARDS[board];
const pages = parseInt(argVal(args, '--pages', '1'), 10);
if (boardId === undefined && board !== 'home')
throw new Error(`未知版块 "${board}",用 --id 指定 forum 编号,或见 -h 的版块列表`);
const urls = [];
for (let p = 1; p <= pages; p++) {
const url = boardId ? `${BASE}/forum-${boardId}${p > 1 ? '-' + p : ''}.htm` : `${BASE}/`;
const html = await fetchText(url);
const items = boardId ? parseForumList(html) : parseKanxueHomeList(html);
urls.push(...items);
}
if (md) {
console.log(`## 看雪 ${board} 文章列表\n`);
for (const it of urls) console.log(`- [${it.title}](${it.url})` + (it.author ? ` — ${it.author}` : ''));
} else {
console.log(JSON.stringify(urls, null, 2));
}
} else if (cmd === 'kanxue' && args[1] === 'thread') {
const id = args[2];
if (!id) throw new Error('缺少帖子ID,如: kanxue thread 292402');
const html = await fetchText(`${BASE}/thread-${id}.htm`);
const t = parseThread(html);
if (md) {
console.log(`# ${t.title}\n\n作者: ${t.author ?? '未知'}\n\n${t.body}`);
} else {
console.log(JSON.stringify({ url: `${BASE}/thread-${id}.htm`, ...t }, null, 2));
}
} else if (cmd === 'wechat') {
const url = args[1];
if (!url) throw new Error('缺少微信文章 URL');
const html = await fetchText(url);
const a = parseWechat(html);
if (md) {
console.log(`# ${a.title ?? '未知'}\n\n作者: ${a.author ?? '未知'}\n\n${a.body}`);
} else {
console.log(JSON.stringify({ url, ...a }, null, 2));
}
} else {
usage();
process.exitCode = 1;
}
}
if (process.argv[1]?.endsWith('wxsource.mjs')) {
main(process.argv).catch((e) => {
console.error(`错误: ${e.message}`);
process.exitCode = 1;
});
}