From 954e8f0ce5fd1f9d051de71697139bcf01f3a109 Mon Sep 17 00:00:00 2001
From: mkaaad <kadmanmk@outlook.com>
Date: Sat, 15 Aug 2026 02:27:08 +0800
Subject: [PATCH 1/2] =?UTF-8?q?feat:=20=E9=A2=91=E9=81=93=E5=8F=B7=20CH-01?=
 =?UTF-8?q?/02=20=E5=9B=9E=E5=88=B0=E5=B7=A6=E5=8F=B3=E7=AE=AD=E5=A4=B4?=
 =?UTF-8?q?=E4=B8=AD=E9=97=B4=EF=BC=88=E5=9B=BA=E5=AE=9A=E5=AE=BD=E5=BA=A6?=
 =?UTF-8?q?=EF=BC=89=EF=BC=8C=E9=A2=91=E9=81=93=E5=90=8D=E4=BF=9D=E7=95=99?=
 =?UTF-8?q?=E5=9C=A8=E5=8F=B3=E4=BE=A7?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 src/components/TvShell.astro | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/src/components/TvShell.astro b/src/components/TvShell.astro
index 52e93b1..82e2e8c 100644
--- a/src/components/TvShell.astro
+++ b/src/components/TvShell.astro
@@ -23,6 +23,7 @@ const BASE = import.meta.env.BASE_URL.replace(/\/$/, '') + '/';
 
   <div class="controls">
     <button class="ctl" id="prev" type="button" aria-label="上一个频道">◀</button>
+    <span class="channel-no" id="channel-no">CH-01</span>
     <button class="ctl" id="next" type="button" aria-label="下一个频道">▶</button>
     <label class="jump">
       跳转到
@@ -187,6 +188,15 @@ const BASE = import.meta.env.BASE_URL.replace(/\/$/, '') + '/';
     opacity: 0.35;
     cursor: not-allowed;
   }
+  .channel-no {
+    flex: 0 0 64px;
+    width: 64px;
+    text-align: center;
+    font-family: ui-monospace, SFMono-Regular, Menlo, monospace;
+    font-size: 14px;
+    letter-spacing: 1px;
+    color: var(--muted);
+  }
   .nav-group {
     display: flex;
     align-items: center;
@@ -289,6 +299,7 @@ const BASE = import.meta.env.BASE_URL.replace(/\/$/, '') + '/';
   const screen = document.getElementById('screen');
   const crt = document.getElementById('crt');
   const crtCh = document.getElementById('crt-ch');
+  const channelNo = document.getElementById('channel-no');
   const chipsEl = document.getElementById('chips');
   const jumpEl = document.getElementById('jump');
   const currentLabel = document.getElementById('current-label');
@@ -320,7 +331,8 @@ const BASE = import.meta.env.BASE_URL.replace(/\/$/, '') + '/';
     setActive(id);
     const target = `${BASE}channels/${id}/`;
     const name = channels[indexOf(id)].name;
-    currentLabel.textContent = `${labelOf(id)} · ${name}`;
+    channelNo.textContent = labelOf(id);
+    currentLabel.textContent = name;
     jumpEl.value = id;
     if (screen.dataset.loaded !== id) {
       crtCh.textContent = labelOf(id);

From 3bdd4eb7fe041c46f6921f97330331526e5a94c3 Mon Sep 17 00:00:00 2001
From: mkaaad <kadmanmk@outlook.com>
Date: Sat, 15 Aug 2026 03:05:08 +0800
Subject: [PATCH 2/2] =?UTF-8?q?feat:=20=E5=AA=92=E4=BD=93=E5=85=A8?=
 =?UTF-8?q?=E9=83=A8=E7=9B=B4=E9=93=BE=E5=8E=9F=E7=AB=99=EF=BC=8C=E9=95=9C?=
 =?UTF-8?q?=E5=83=8F=E4=BB=8E=20~405MB=20=E7=98=A6=E8=BA=AB=E5=88=B0=20~30?=
 =?UTF-8?q?MB=EF=BC=8C=E6=96=B0=E5=A2=9E=20sfkm=20=E9=A2=91=E9=81=93?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- wget 不再下载媒体/字体（reject-regex + 文件魔数兜底无扩展名图片）
- 清洗时图片/视频/音频/字体引用一律改写为原站绝对 URL，CSS/JS 仍本地化
- CSP 放开 img/media/font 的 https，修复 met6 友链页外站头像被拦为空的问题
- 移除 extraAssets 下载机制，rycarl 动态壁纸视频一并直链 image.rycarl.cn
- 缓存 key 换前缀作废旧的大缓存，增量/删除同步逻辑不变
- 新增 blog.sfkm.me（Snowykami's Blog）频道
---
 .github/workflows/mirror.yml |   5 +-
 README.md                    |   6 +-
 scripts/mirror.mjs           |  71 ++++-----
 scripts/sanitize.mjs         | 273 +++++++++++++++++++++++++++--------
 src/data/channels.json       |  13 +-
 5 files changed, 265 insertions(+), 103 deletions(-)

diff --git a/.github/workflows/mirror.yml b/.github/workflows/mirror.yml
index add0add..e14d3be 100644
--- a/.github/workflows/mirror.yml
+++ b/.github/workflows/mirror.yml
@@ -35,9 +35,10 @@ jobs:
           path: .cache
           # 稳定 key：每次成功运行覆盖同一个缓存条目，避免堆积 370MB 级快照，
           # 并保证增量状态（.cache/state）在下次运行一定能恢复
-          key: mirror-${{ hashFiles('src/data/channels.json') }}
+          # 媒体改为直链后缓存只有 HTML/CSS/JS，换前缀作废旧的大缓存
+          key: mirror-hotlink-${{ hashFiles('src/data/channels.json') }}
           restore-keys: |
-            mirror-
+            mirror-hotlink-
 
       - name: Mirror and sanitize
         run: node scripts/mirror.mjs
diff --git a/README.md b/README.md
index 275ca7b..acbdbf3 100644
--- a/README.md
+++ b/README.md
@@ -13,7 +13,10 @@ npm run dev          # 本地预览
 
 本地快速验证可加环境变量：`MIRROR_QUICK=1 npm run mirror`（只抓两层页面）。
 
-镜像默认不下载 Halo 的 `?width=` 缩略图变体（静态托管下无收益），页面 `srcset` 会自动收敛到原图；多频道并发抓取（最多 3 路）。
+镜像只下载 HTML/CSS/JS 及少量非媒体附件（zip/pdf/xml），图片、视频、音频、字体一律不落盘，
+清洗时把引用改写为原站绝对 URL（直链），因此镜像体积小（总量约 28MB 量级）、CI 拉取快；
+代价是访客浏览时由浏览器直接向友站请求媒体流量，友站若有防盗链（Referer 校验）可能破图。
+多频道并发抓取（最多 3 路）。
 
 增量同步：有状态后按 `sitemap.xml` 的 `lastmod` 对比，只拉新增/更新的文章页，源站删除的页面会同步删除；首页与归档列表始终刷新；首次运行或 sitemap 不可用时自动回退全量抓取。状态文件在 `.cache/state/`，随 Actions 缓存保留。
 
@@ -35,7 +38,6 @@ npm run dev          # 本地预览
 
 - `extraOrigins`：图片放在独立图床时，把图床域名加进 `img-src` 白名单，如 `["https://image.example.com:443"]`。
 - `alwaysFetch`：sitemap 未收录但希望始终刷新的页面，如 `["/links"]`。
-- `extraAssets`：希望镜像到本地的外站资源（如动态壁纸视频），`url` 为原地址、`path` 为镜像内相对路径；已存在时跳过，随 CI 缓存保留。
 
 提交 PR 即视为同意本站对博客做静态镜像（保留署名、功能引导回原站）。CI 会自动校验格式。
 
diff --git a/scripts/mirror.mjs b/scripts/mirror.mjs
index 3cd7c53..eeb65bb 100644
--- a/scripts/mirror.mjs
+++ b/scripts/mirror.mjs
@@ -4,7 +4,7 @@ import { existsSync } from 'node:fs';
 import path from 'node:path';
 import { fileURLToPath } from 'node:url';
 import astroConfig from '../astro.config.mjs';
-import { sanitizeChannel } from './sanitize.mjs';
+import { sanitizeChannel, MEDIA_EXT_RE, SAFE_FILE_RE, looksLikeMedia } from './sanitize.mjs';
 
 const ROOT = path.resolve(path.dirname(fileURLToPath(import.meta.url)), '..');
 const base = (astroConfig.base ?? '').replace(/\/$/, '');
@@ -13,6 +13,10 @@ const outRoot = path.join(ROOT, 'public/channels');
 const stateRoot = path.join(ROOT, '.cache/state');
 // 首页与归档列表不在 sitemap 中，但新文章会出现在这里，始终刷新
 const ALWAYS_FETCH = ['/', '/archives'];
+// 媒体/字体不再下载：wget 直接拒绝（含 ?width= 缩略图变体），
+// 清洗阶段再把引用直链回原站。zip/pdf/xml 等非媒体附件仍下载。
+const REJECT_MEDIA_RE =
+  '(\\?width=)|\\.(png|jpe?g|webp|gif|svg|ico|avif|bmp|apng|mp4|webm|ogv|mov|mp3|wav|ogg|oga|flac|aac|m4a|woff2?|ttf|otf|eot)([?#]|$)';
 
 const channelsData = JSON.parse(await readFile(path.join(ROOT, 'src/data/channels.json'), 'utf8'));
 const channels = (channelsData.channels ?? []).filter((c) => c.enabled !== false);
@@ -39,8 +43,8 @@ async function processChannel(channel) {
   // 缓存目录为空（首次 / 缓存丢失）时，先全量抓取并播种状态
   if (!hostDir) {
     const args = [
-      '-m', '-p', '-np', '-E', '-k', '-nv', '--no-remove-listing',
-      '--timeout=30', '--tries=3', '--reject-regex', '\\?width=',
+      '-m', '-p', '-np', '-E', '-nv', '--no-remove-listing',
+      '--timeout=30', '--tries=3', '--reject-regex', REJECT_MEDIA_RE,
       '-P', cacheDir, channel.url,
     ];
     if (process.env.MIRROR_QUICK) args.push('--level=2');
@@ -78,8 +82,6 @@ async function processChannel(channel) {
     }
   }
 
-  await fetchExtraAssets(channel, hostDir);
-
   try {
     await sanitizeChannel({
       hostDir,
@@ -89,8 +91,9 @@ async function processChannel(channel) {
       prefix,
       base: base || '',
       extraOrigins: channel.extraOrigins || [],
-      extraAssets: channel.extraAssets || [],
     });
+    // 镜像产物已无媒体；顺手清掉缓存里的旧媒体文件，避免 Actions 缓存滞留
+    await purgeMediaFiles(hostDir);
     const { sizeBytes, fileCount } = await dirStats(outDir);
     entry.sizeBytes = sizeBytes;
     entry.fileCount = fileCount;
@@ -106,31 +109,6 @@ async function processChannel(channel) {
   return entry;
 }
 
-// 把配置里声明的外站资源（如动态壁纸视频）镜像到本地，避免访客每次访问都请求友站图床。
-// 只下载一次：文件已存在则跳过，随 Actions 缓存保留。
-async function fetchExtraAssets(channel, hostDir) {
-  if (!hostDir) return;
-  for (const asset of channel.extraAssets || []) {
-    if (!asset?.url || !asset?.path || asset.path.includes('..')) continue;
-    const target = path.join(hostDir, asset.path);
-    if (existsSync(target)) continue;
-    await mkdir(path.dirname(target), { recursive: true });
-    console.log(`[${channel.id}] 下载额外资源 ${asset.url}`);
-    try {
-      const res = await fetch(asset.url, { signal: AbortSignal.timeout(120000) });
-      if (!res.ok) {
-        console.warn(`[${channel.id}] 额外资源下载失败 HTTP ${res.status}：${asset.url}`);
-        continue;
-      }
-      const buf = Buffer.from(await res.arrayBuffer());
-      await writeFile(target, buf);
-      console.log(`[${channel.id}] 额外资源已保存 ${asset.path}（${(buf.length / 1048576).toFixed(1)} MB）`);
-    } catch (e) {
-      console.warn(`[${channel.id}] 额外资源下载失败：${asset.url} (${e.message})`);
-    }
-  }
-}
-
 // 增量更新：对照 sitemap + 状态文件，只拉新增/更新的页面，并删除源站已移除的页面
 async function tryIncremental(channel, cacheDir) {
   const stateFile = path.join(stateRoot, `${channel.id}.json`);
@@ -186,8 +164,8 @@ async function tryIncremental(channel, cacheDir) {
   if (urls.size) {
     fetched = urls.size;
     const args = [
-      '-E', '-k', '-p', '-np', '-N', '-nv', '--timeout=30', '--tries=3',
-      '--reject-regex', '\\?width=', '-P', cacheDir, ...urls,
+      '-E', '-p', '-np', '-N', '-nv', '--timeout=30', '--tries=3',
+      '--reject-regex', REJECT_MEDIA_RE, '-P', cacheDir, ...urls,
     ];
     const res = spawnSync('wget', args, { stdio: 'inherit' });
     if (res.status !== 0) {
@@ -389,3 +367,30 @@ async function looksLikeHtml(file) {
     return false;
   }
 }
+
+async function listAllFiles(dir) {
+  const out = [];
+  async function walk(d) {
+    for (const entry of await readdir(d, { withFileTypes: true })) {
+      const p = path.join(d, entry.name);
+      if (entry.isDirectory()) await walk(p);
+      else if (entry.isFile()) out.push(p);
+    }
+  }
+  await walk(dir);
+  return out;
+}
+
+// 删除缓存中遗留的媒体/字体文件（wget 新拉取已按 REJECT_MEDIA_RE 拒绝）
+async function purgeMediaFiles(dir) {
+  if (!existsSync(dir)) return;
+  let removed = 0;
+  for (const file of await listAllFiles(dir)) {
+    const rel = path.relative(dir, file);
+    if (MEDIA_EXT_RE.test(rel) || (!SAFE_FILE_RE.test(rel) && (await looksLikeMedia(file)))) {
+      await rm(file, { force: true });
+      removed += 1;
+    }
+  }
+  if (removed) console.log(`[cache] 清理 ${removed} 个媒体/字体缓存文件`);
+}
diff --git a/scripts/sanitize.mjs b/scripts/sanitize.mjs
index afb9b06..4d54d0c 100644
--- a/scripts/sanitize.mjs
+++ b/scripts/sanitize.mjs
@@ -9,21 +9,109 @@ const REMOVE_SCRIPT_RE =
 const INLINE_REMOVE_RE = /\/plugins\/(PluginCommentWidget|PluginSearchWidget)/;
 const KNOWN_ROOT_RE =
   /(["'])\/(upload|themes|plugins|archives|tags|categories|links|pages|images|search|rss|sitemap|favicon|about-me|met-website)\/([^"'\s\\]*)/g;
-// 部分友站会把图片/头像放在独立图床，镜像不抓跨域资源，
-// 需要把这些来源加进 img-src 白名单，页面才能正常显示图片。
+
+// 媒体/字体文件扩展名：这些资源不再下载，引用一律直链原站绝对 URL
+export const MEDIA_EXT_RE =
+  /\.(?:png|jpe?g|webp|gif|svg|ico|avif|bmp|apng|mp4|webm|ogv|mov|mp3|wav|ogg|oga|flac|aac|m4a|woff2?|ttf|otf|eot)(?:[?#]|$)/i;
+
+// 已知非媒体扩展名：这类文件不做魔数探测（避免误删 sitemap.xml 等文本文件）
+export const SAFE_FILE_RE =
+  /\.(?:html?|css|js|json|xml|txt|md|yml|yaml|map|webmanifest|gz|zip|pdf)(?:[?#]|$)/i;
+
+// 部分友站用无扩展名的 URL 提供图片/视频（如 /api/v1/file/f/1），
+// 按文件头魔数识别并删除，通用兜底所有站点。
+const MEDIA_MAGIC = [
+  [0x89, 0x50, 0x4e, 0x47], // PNG
+  [0xff, 0xd8, 0xff], // JPEG
+  [0x47, 0x49, 0x46, 0x38], // GIF
+  [0x52, 0x49, 0x46, 0x46], // RIFF（WEBP/AVIF）
+  [0x1a, 0x45, 0xdf, 0xa3], // WebM/Matroska
+  [0x49, 0x44, 0x33], // MP3（ID3）
+  [0x4f, 0x67, 0x67, 0x53], // Ogg
+  [0x00, 0x01, 0x00, 0x00], // TTF
+  [0x4f, 0x54, 0x54, 0x4f], // OTF（OTTO）
+  [0x00, 0x00, 0x01, 0x00], // ICO
+  [0x42, 0x4d], // BMP
+  [0x49, 0x49, 0x2a, 0x00], // TIFF LE
+  [0x4d, 0x4d, 0x00, 0x2a], // TIFF BE
+  [0x77, 0x4f, 0x46, 0x46], // WOFF
+  [0x77, 0x4f, 0x46, 0x32], // WOFF2
+];
+
+export async function looksLikeMedia(file) {
+  try {
+    const fh = await open(file, 'r');
+    const buf = Buffer.alloc(32);
+    const { bytesRead } = await fh.read(buf, 0, 32, 0);
+    await fh.close();
+    const b = buf.subarray(0, bytesRead);
+    if (bytesRead < 4) return false;
+    for (const magic of MEDIA_MAGIC) {
+      if (b.length >= magic.length && magic.every((byte, i) => b[i] === byte)) return true;
+    }
+    // MP4/M4V：偏移 4 处为 ftyp 魔数（盒子大小可变，不能用固定字节序列）
+    if (b.length >= 8 && b.subarray(4, 8).toString('latin1') === 'ftyp') return true;
+    const head = b.toString('latin1').trimStart();
+    return /^<svg[\s>]/i.test(head);
+  } catch {
+    return false;
+  }
+}
+
+// 这些标签的 src 属性内容必定是媒体（即使 URL 无扩展名，如 Next.js /_next/image?...）
+const MEDIA_SRC_TAGS = new Set(['img', 'video', 'audio', 'source', 'embed', 'track', 'image']);
+
+function isMediaRef(value) {
+  if (typeof value !== 'string') return false;
+  try {
+    return MEDIA_EXT_RE.test(new URL(value, 'http://x').pathname);
+  } catch {
+    return false;
+  }
+}
+
+function originNoSlash(origin) {
+  return (origin || '').replace(/\/$/, '');
+}
+
+// 镜像内相对路径 -> 原站目录（用于把相对媒体引用解析回原站绝对 URL）
+function originDirOf(rel) {
+  const dir = path.posix.dirname(rel.split(path.sep).join('/'));
+  return dir === '.' ? '/' : `/${dir}/`;
+}
+
 function buildCsp(extraOrigins = []) {
-  const img = ["'self'", "'data:'", ...extraOrigins].join(' ');
-  const media = ["'self'", ...extraOrigins].join(' ');
-  return `default-src 'self'; script-src 'self' 'unsafe-inline' 'unsafe-eval'; style-src 'self' 'unsafe-inline'; img-src ${img}; font-src 'self' data:; media-src ${media}; connect-src 'none'; frame-src 'none'; object-src 'none'; base-uri 'self'; form-action 'self'`;
+  // 媒体不再镜像，img/video/audio/font 一律直链原站，放开 https；
+  // 脚本/连接仍保持收紧（script-src 'self'、connect-src 'none'）
+  const img = ["'self'", "'data:'", "https:", ...extraOrigins].join(' ');
+  const media = ["'self'", "https:", ...extraOrigins].join(' ');
+  const font = ["'self'", "data:", "https:", ...extraOrigins].join(' ');
+  return `default-src 'self'; script-src 'self' 'unsafe-inline' 'unsafe-eval'; style-src 'self' 'unsafe-inline'; img-src ${img}; font-src ${font}; media-src ${media}; connect-src 'none'; frame-src 'none'; object-src 'none'; base-uri 'self'; form-action 'self'`;
 }
 
 function escapeRegExp(s) {
   return s.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
 }
 
-function rewriteRoot(value, prefix, origin) {
+function rewriteRoot(value, prefix, origin, { media = false, baseDir = '/' } = {}) {
   if (typeof value !== 'string') return value;
   const clean = value.replace(/\?width=[^&]*/, '');
+  // 媒体引用：一律改写为原站绝对 URL（根相对/协议相对/站内相对 → 绝对；绝对 URL 原样保留）
+  if (media) {
+    if (clean.startsWith('/') && !clean.startsWith('//')) {
+      return `${originNoSlash(origin)}${clean}`;
+    }
+    if (clean.startsWith('//')) return `https:${clean}`;
+    if (/^[a-z][a-z0-9+.-]*:/i.test(clean)) return clean;
+    if (origin) {
+      try {
+        return new URL(clean, `${originNoSlash(origin)}${baseDir}`).href;
+      } catch {
+        return clean;
+      }
+    }
+    return clean;
+  }
   if (clean.startsWith('/') && !clean.startsWith('//')) {
     return prefix + clean;
   }
@@ -49,11 +137,12 @@ function rewriteAssetExt(value) {
     : value;
 }
 
-function cleanRef(value, prefix, origin) {
+function cleanRef(value, prefix, origin, opts = {}) {
   if (typeof value !== 'string') return value;
-  let v = rewriteAssetExt(rewriteRoot(value, prefix, origin));
+  let v = rewriteAssetExt(rewriteRoot(value, prefix, origin, opts));
   // wget 把含查询串的资源存成字面 "?":main.css?v=1.0.6，引用需编码为 %3F，
-  // 否则浏览器把 ? 当查询串、按 main.css 找文件 → 404（页面白屏）
+  // 否则浏览器把 ? 当查询串、按 main.css 找文件 → 404（页面白屏）。
+  // 绝对 URL（含媒体直链）保持真实查询串，不编码。
   if (!/^(https?:)?\/\//i.test(v)) v = v.replace(/\?/g, '%3F');
   return v;
 }
@@ -77,13 +166,13 @@ function resolveLocalPath(pathname, outDir) {
   return null;
 }
 
-function rewriteSrcset(value, prefix, origin) {
+function rewriteSrcset(value, prefix, origin, baseDir) {
   return value
     .split(',')
     .map((part) => {
       const m = part.trim().match(/^(\S+)(\s+.+)?$/);
       if (!m) return part;
-      return cleanRef(m[1], prefix, origin) + (m[2] ?? '');
+      return cleanRef(m[1], prefix, origin, { media: true, baseDir }) + (m[2] ?? '');
     })
     .join(', ');
 }
@@ -113,32 +202,58 @@ async function looksLikeHtml(file) {
   }
 }
 
-function rewriteJsonStrings(node, prefix, origin) {
+function rewriteJsonStrings(node, prefix, origin, baseDir) {
   if (typeof node === 'string') {
-    return cleanRef(node, prefix, origin);
+    return cleanRef(node, prefix, origin, { media: isMediaRef(node), baseDir });
   }
   if (Array.isArray(node)) {
-    return node.map((v) => rewriteJsonStrings(v, prefix, origin));
+    return node.map((v) => rewriteJsonStrings(v, prefix, origin, baseDir));
   }
   if (node && typeof node === 'object') {
     const out = {};
-    for (const [k, v] of Object.entries(node)) out[k] = rewriteJsonStrings(v, prefix, origin);
+    for (const [k, v] of Object.entries(node)) out[k] = rewriteJsonStrings(v, prefix, origin, baseDir);
     return out;
   }
   return node;
 }
 
-function rewriteCssUrls(css, prefix, origin) {
-  // 根相对路径 url(/upload/...)
-  css = css.replace(/(url\(\s*['"]?)\//g, `$1${prefix}/`);
-  // 同源绝对路径 url(https://原站/upload/...)
-  if (origin) {
-    const o = origin.replace(/\/$/, '');
-    css = css.replace(new RegExp(`(url\\(\\s*['"]?)${escapeRegExp(o)}`, 'g'), `$1${prefix}`);
-  }
-  // 去掉 .gz 后缀的引用（镜像内只有未压缩版）
-  css = css.replace(/(url\(\s*['"]?)([^)'"]+)/g, (m, pre, p) => pre + rewriteAssetExt(p));
-  return css;
+function rewriteCssUrls(css, prefix, origin, cssDir = '/') {
+  const o = originNoSlash(origin);
+  return css.replace(/url\(\s*['"]?([^'")]+)['"]?\s*\)/g, (m, p) => {
+    const target = p.trim();
+    // data: 与函数式 url() 原样保留
+    if (/^data:/i.test(target) || target.includes('(')) return m;
+    const media = isMediaRef(target);
+    if (target.startsWith('/') && !target.startsWith('//')) {
+      return `url(${media ? o + target : prefix + target})`;
+    }
+    if (target.startsWith('//')) {
+      return `url(${media ? 'https:' + target : target})`;
+    }
+    if (/^https?:/i.test(target)) {
+      if (origin) {
+        try {
+          const u = new URL(target, origin);
+          if (u.origin === new URL(origin).origin) {
+            return `url(${media ? target : prefix + u.pathname + (u.search || '')})`;
+          }
+        } catch {
+          // 解析失败则原样保留
+        }
+      }
+      return `url(${target})`;
+    }
+    // 相对路径：媒体按 CSS 所在站内目录解析为原站绝对 URL；资源保持相对（本地文件同构）
+    if (media && origin) {
+      try {
+        return `url(${new URL(target, o + cssDir).href})`;
+      } catch {
+        return `url(${target})`;
+      }
+    }
+    // 去掉 .gz 后缀的引用（镜像内只有未压缩版）
+    return `url(${rewriteAssetExt(target)})`;
+  });
 }
 
 // 部分友站对 CSS/JS 下发的是 gzip 字节（Content-Encoding: gzip），
@@ -195,7 +310,7 @@ async function listFiles(dir) {
   return out;
 }
 
-async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins = [], assetMap = {} }) {
+async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins = [], pageDir = '/' }) {
   let html = await readFile(file, 'utf8');
   const $ = load(html, { decodeEntities: false });
   let faviconPath = null;
@@ -218,17 +333,32 @@ async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins =
 
   $('[href]').each((_, el) => {
     const v = $(el).attr('href');
-    // <a> 链接由下方专门的同源解析逻辑处理，避免把外站/未镜像链接误改成本地 404
-    const nv = cleanRef(v, prefix, (el.tagName || '').toLowerCase() === 'a' ? null : origin);
+    const tag = (el.tagName || '').toLowerCase();
+    if (tag === 'a') {
+      // <a> 链接由下方专门的同源解析逻辑处理；指向媒体文件的链接直链原站
+      if (isMediaRef(v)) {
+        const nv = cleanRef(v, prefix, origin, { media: true, baseDir: pageDir });
+        if (nv !== v) $(el).attr('href', nv);
+      } else {
+        const nv = cleanRef(v, prefix, null);
+        if (nv !== v) $(el).attr('href', nv);
+      }
+      return;
+    }
+    const rel = ($(el).attr('rel') || '').toLowerCase();
+    const media = (tag === 'link' && rel.includes('icon')) || isMediaRef(v);
+    const nv = cleanRef(v, prefix, origin, { media, baseDir: pageDir });
     if (nv !== v) $(el).attr('href', nv);
   });
   $('[src]').each((_, el) => {
     const v = $(el).attr('src');
-    const nv = cleanRef(v, prefix, origin);
+    const tag = (el.tagName || '').toLowerCase();
+    const media = MEDIA_SRC_TAGS.has(tag) || isMediaRef(v);
+    const nv = cleanRef(v, prefix, origin, { media, baseDir: pageDir });
     if (nv !== v) $(el).attr('src', nv);
     // 音乐组件等会把专辑封面误写成页面地址（xxx.html），
     // 转成透明占位图，避免满页破图图标
-    if ((el.tagName || '').toLowerCase() === 'img' && /\.html?$/i.test(nv || '')) {
+    if (tag === 'img' && /\.html?$/i.test(nv || '')) {
       $(el).attr(
         'src',
         'data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7',
@@ -236,7 +366,12 @@ async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins =
     }
   });
   $('[srcset]').each((_, el) => {
-    $(el).attr('srcset', rewriteSrcset($(el).attr('srcset'), prefix, origin));
+    $(el).attr('srcset', rewriteSrcset($(el).attr('srcset'), prefix, origin, pageDir));
+  });
+  $('[poster]').each((_, el) => {
+    const v = $(el).attr('poster');
+    const nv = cleanRef(v, prefix, origin, { media: true, baseDir: pageDir });
+    if (nv !== v) $(el).attr('poster', nv);
   });
   // 同源绝对链接改写为镜像本地路径（点击后仍在 iframe 内导航）；
   // 外站或找不到本地文件的链接一律新标签打开，避免跨域导航被 X-Frame-Options
@@ -259,6 +394,13 @@ async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins =
       return;
     }
     if (url.origin === channelOrigin) {
+      // 媒体文件（图片/视频等）未镜像到本地，直链原站并在新标签打开，避免本地 404
+      if (isMediaRef(url.pathname)) {
+        $(el).attr('href', url.href);
+        $(el).attr('target', '_blank');
+        $(el).attr('rel', 'noopener');
+        return;
+      }
       const resolved = resolveLocalPath(url.pathname, outDir);
       if (resolved) {
         $(el).attr('href', `${prefix}/${resolved}`);
@@ -270,7 +412,8 @@ async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins =
   });
   $('[data-src]').each((_, el) => {
     const v = $(el).attr('data-src');
-    const nv = cleanRef(v, prefix, origin);
+    const tag = (el.tagName || '').toLowerCase();
+    const nv = cleanRef(v, prefix, origin, { media: tag === 'img' || isMediaRef(v), baseDir: pageDir });
     if (nv !== v) $(el).attr('data-src', nv);
   });
   $('[action]').each((_, el) => {
@@ -280,12 +423,20 @@ async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins =
   });
   $('meta[property^="og:"], meta[name^="twitter:"]').each((_, el) => {
     const v = $(el).attr('content');
-    const nv = cleanRef(v, prefix, origin);
+    const key = ($(el).attr('property') || $(el).attr('name') || '').toLowerCase();
+    const media =
+      key === 'og:image' ||
+      key === 'og:image:url' ||
+      key === 'og:image:secure_url' ||
+      key === 'twitter:image' ||
+      key === 'twitter:image:src' ||
+      isMediaRef(v);
+    const nv = cleanRef(v, prefix, origin, { media, baseDir: pageDir });
     if (nv !== v) $(el).attr('content', nv);
   });
   $('[style]').each((_, el) => {
     const v = $(el).attr('style') || '';
-    const nv = rewriteCssUrls(v, prefix, origin);
+    const nv = rewriteCssUrls(v, prefix, origin, pageDir);
     if (nv !== v) $(el).attr('style', nv);
   });
 
@@ -299,7 +450,7 @@ async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins =
   $('script[type="application/json"]').each((_, el) => {
     const raw = $(el).text();
     try {
-      const data = rewriteJsonStrings(JSON.parse(raw), prefix, origin);
+      const data = rewriteJsonStrings(JSON.parse(raw), prefix, origin, pageDir);
       $(el).text(JSON.stringify(data));
     } catch {
       // 非 JSON 原样保留
@@ -316,23 +467,24 @@ async function sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins =
   }
 
   html = $.html();
-  // 镜像到本地的外站资源（如动态壁纸视频）：把原站地址精确替换为本地副本，
-  // 保持原页面结构与主题配置不变（type 仍为 video，<video> 元素原样保留）
-  for (const [from, to] of Object.entries(assetMap)) {
-    if (html.includes(from)) html = html.split(from).join(to);
-  }
   // 处理内联 <script> 中带引号的根绝对路径（themeConfig 等），
   // 只改脚本内容（保留 <script> 外壳），避免误伤文章正文里的路径示例
   html = html.replace(/<script\b([^>]*)>([\s\S]*?)<\/script>/gi, (m, attrs, body) =>
-    `<script${attrs}>${body.replace(KNOWN_ROOT_RE, (sub, q, dir, rest) => `${q}${prefix}/${dir}/${rewriteAssetExt(rest)}`)}</script>`,
+    `<script${attrs}>${body.replace(
+      KNOWN_ROOT_RE,
+      (sub, q, dir, rest) =>
+        isMediaRef(`/${dir}/${rest}`)
+          ? `${q}${originNoSlash(origin)}/${dir}/${rest.replace(/\?width=[^&]*/, '')}`
+          : `${q}${prefix}/${dir}/${rewriteAssetExt(rest)}`,
+    )}</script>`,
   );
   await writeFile(file, html, 'utf8');
   return faviconPath;
 }
 
-async function sanitizeCss(file, prefix, origin) {
+async function sanitizeCss(file, prefix, origin, cssDir) {
   let css = await readFile(file, 'utf8');
-  css = rewriteCssUrls(css, prefix, origin);
+  css = rewriteCssUrls(css, prefix, origin, cssDir);
   await writeFile(file, css, 'utf8');
 }
 
@@ -340,11 +492,17 @@ async function sanitizeJs(file, prefix, origin) {
   let js = await readFile(file, 'utf8');
   js = js.replace(
     KNOWN_ROOT_RE,
-    (m, q, dir, rest) => `${q}${prefix}/${dir}/${rewriteAssetExt(rest)}`,
+    (m, q, dir, rest) =>
+      isMediaRef(`/${dir}/${rest}`)
+        ? `${q}${originNoSlash(origin)}/${dir}/${rest.replace(/\?width=[^&]*/, '')}`
+        : `${q}${prefix}/${dir}/${rewriteAssetExt(rest)}`,
   );
-  // 内联/外链 JS 里的同源绝对地址（如 themeConfig 中的 /upload/...）也改为本地路径
+  // 内联/外链 JS 里的同源绝对地址：媒体保持原站直链，资源（CSS/JS/页面）改为本地路径
   if (origin) {
-    js = js.split(origin.replace(/\/$/, '')).join(prefix);
+    const o = originNoSlash(origin);
+    js = js.replace(new RegExp(escapeRegExp(o) + '(/[^"\'\\s\\\\]*)', 'g'), (m, p) =>
+      isMediaRef(p) ? m : prefix + p,
+    );
   }
   await writeFile(file, js, 'utf8');
 }
@@ -357,41 +515,36 @@ export async function sanitizeChannel({
   prefix,
   base,
   extraOrigins = [],
-  extraAssets = [],
 }) {
   await rm(outDir, { recursive: true, force: true });
   await mkdir(outDir, { recursive: true });
   await copyTree(hostDir, outDir);
   await decompressGzAssets(outDir);
 
-  // 已成功镜像到本地的外站资源：url -> 本地路径（只映射真实存在的文件）
-  const assetMap = {};
-  for (const asset of extraAssets) {
-    if (!asset?.url || !asset?.path || asset.path.includes('..')) continue;
-    if (existsSync(path.join(outDir, asset.path))) {
-      assetMap[asset.url] = `${prefix}/${asset.path}`;
-    }
-  }
-
   const files = await listFiles(outDir);
   let faviconPath = null;
 
   for (const file of files) {
     const rel = path.relative(outDir, file);
-    if (rel.includes('?width=')) {
+    if (
+      rel.includes('?width=') ||
+      MEDIA_EXT_RE.test(rel) ||
+      (!SAFE_FILE_RE.test(rel) && (await looksLikeMedia(file)))
+    ) {
       await rm(file, { force: true });
       continue;
     }
+    const pageDir = originDirOf(rel);
     const isHtml =
       rel.endsWith('.html') ||
       (!/\.(css|js|png|jpe?g|webp|gif|svg|ico|woff2?|ttf|otf|eot|json|xml|txt|gz|zip|pdf|mp3|mp4|webm)$/i.test(rel) &&
         (await looksLikeHtml(file)));
     if (isHtml) {
       faviconPath =
-        (await sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins, assetMap })) ||
+        (await sanitizeHtml(file, { origin, prefix, base, outDir, extraOrigins, pageDir })) ||
         faviconPath;
     } else if (rel.endsWith('.css')) {
-      await sanitizeCss(file, prefix, origin);
+      await sanitizeCss(file, prefix, origin, pageDir);
     } else if (rel.endsWith('.js')) {
       await sanitizeJs(file, prefix, origin);
     }
diff --git a/src/data/channels.json b/src/data/channels.json
index 2b421e0..0b099f0 100644
--- a/src/data/channels.json
+++ b/src/data/channels.json
@@ -14,12 +14,13 @@
       "description": "Rycarl 的小博客",
       "extraOrigins": ["https://image.rycarl.cn:443"],
       "alwaysFetch": ["/links"],
-      "extraAssets": [
-        {
-          "url": "https://image.rycarl.cn:443/pic/compressed-compressed-炎夏凉梦-横屏4K.mp4",
-          "path": "upload/dynamic-wallpaper.mp4"
-        }
-      ],
+      "enabled": true
+    },
+    {
+      "id": "sfkm",
+      "name": "Snowykami's Blog",
+      "url": "https://blog.sfkm.me/",
+      "description": "SnowyKami 的博客（neo-blog）",
       "enabled": true
     }
   ]