elanous 0.2.9 → 0.2.11

package/skills/youtube-master/scripts/main.ts

@@ -0,0 +1,435 @@
+#!/usr/bin/env -S npx tsx
+import { parseArgs } from 'node:util';
+import { mkdtemp, rm, readFile } from 'node:fs/promises';
+import { join } from 'node:path';
+import { tmpdir } from 'node:os';
+
+import { initEnv, env } from '../src/env.js';
+import { extractVideoId, fetchYoutubeMeta, formatTime } from '../src/youtube.js';
+import { fetchSupadataTranscript, judgeTranscriptSufficiency } from '../src/transcript.js';
+import { downloadAudio, splitAudio } from '../src/audio.js';
+import { transcribeChunks } from '../src/stt.js';
+import { summarize, parseMetaFromSummary } from '../src/summarize.js';
+import { getSaveDir, saveMarkdown, prepareRunDir, persistArtifacts } from '../src/obsidian.js';
+import { classifyDomain, stagingEnabled, stagingSubdir } from '../src/domain-staging.js';
+import { buildStudyNote } from '../src/study-note.js';
+import { decideRoute } from '../src/router.js';
+import type { VideoMeta, ArtifactInfo, RouteDecision, TranscriptSegment, OutputFormat } from '../src/types.js';
+
+// ── Init ──
+
+initEnv();
+
+const { values: flags, positionals } = parseArgs({
+  allowPositionals: true,
+  options: {
+    message:          { type: 'string', short: 'm' },
+    intent:           { type: 'string' },
+    only:             { type: 'string' },
+    format:           { type: 'string' },
+    target:           { type: 'string' },
+    'cloud-stt':      { type: 'boolean', default: false },
+    'transcript-file': { type: 'string' },
+    print:            { type: 'boolean', default: false },
+    'dry-run':        { type: 'boolean', default: false },
+    'no-obsidian':    { type: 'boolean', default: false },
+    'output-dir':     { type: 'string' },
+    'self-test':      { type: 'boolean', default: false },
+    help:             { type: 'boolean', short: 'h', default: false },
+  },
+});
+
+if (flags.help) {
+  printHelp();
+  process.exit(0);
+}
+
+if (flags['self-test']) {
+  runSelfTest();
+  process.exit(0);
+}
+
+const videoUrl = positionals[0];
+if (!videoUrl) {
+  console.error('Error: YouTube URL을 입력해주세요.');
+  printHelp();
+  process.exit(1);
+}
+
+main().catch((err) => {
+  console.error(`Error: ${err.message}`);
+  process.exit(1);
+});
+
+// ── Main pipeline ──
+
+async function main() {
+  const videoId = extractVideoId(videoUrl);
+  if (!videoId) {
+    throw new Error('유효하지 않은 YouTube URL입니다.');
+  }
+
+  const intentText = [flags.intent, flags.message, ...positionals.slice(1)].filter(Boolean).join(' ').trim();
+
+  // Fetch metadata (needed for routing and output)
+  let meta: VideoMeta | null = null;
+  try {
+    console.log('  메타데이터 조회...');
+    meta = await fetchYoutubeMeta(videoId);
+    console.log(`  ${meta.title} (${meta.channel}, ${meta.duration})`);
+  } catch (e: any) {
+    console.log(`  메타데이터 조회 실패: ${e.message}`);
+  }
+
+  // Route decision
+  const route = decideRoute({
+    intentText,
+    durationSec: meta?.durationSec,
+    only: flags.only,
+    forceFormat: flags.format,
+    forceTarget: flags['no-obsidian'] ? 'markdown' : flags.target,
+    cloudStt: flags['cloud-stt'],
+  });
+
+  console.log(`  라우트: format=${route.format}, target=${route.target}, transcript=${route.transcriptStrategy}`);
+  console.log(`  이유: ${route.reason}`);
+
+  if (flags['dry-run']) {
+    console.log(JSON.stringify({ route, videoId, title: meta?.title, durationSec: meta?.durationSec }, null, 2));
+    return;
+  }
+
+  // ── Dispatch by format ──
+
+  if (route.format === 'metadata') {
+    console.log(JSON.stringify(meta, null, 2));
+    return;
+  }
+
+  if (route.format === 'transcript') {
+    const transcript = await acquireTranscript(videoUrl, meta, route, videoId);
+    if (flags.print) {
+      console.log('\n---BEGIN_YOUTUBE_MASTER_MARKDOWN---\n');
+      console.log(transcript);
+      console.log('\n---END_YOUTUBE_MASTER_MARKDOWN---\n');
+    } else {
+      console.log(transcript);
+    }
+    return;
+  }
+
+  // Summary formats: brief, cards, detailed, study-note
+  const summaryFormat: 'brief' | 'cards' | 'detailed' =
+    route.format === 'study-note' ? 'cards' :
+    (['brief', 'cards', 'detailed'].includes(route.format) ? route.format as 'brief' | 'cards' | 'detailed' : 'cards');
+
+  // Acquire transcript
+  const transcript = await acquireTranscript(videoUrl, meta, route, videoId);
+
+  // Build timeline hint (for cards/detailed)
+  let timelineHint = '';
+  if (summaryFormat !== 'brief' && (globalSegments?.length ?? 0) > 0) {
+    const interval = 30;
+    let nextMark = 0;
+    const samples: string[] = [];
+    for (const seg of globalSegments!) {
+      if (seg.offset >= nextMark) {
+        const secs = Math.floor(seg.offset);
+        samples.push(`[${formatTime(seg.offset)}](https://www.youtube.com/watch?v=${videoId}&t=${secs}s) ${seg.text}`);
+        nextMark = seg.offset + interval;
+      }
+    }
+    timelineHint = samples.join('\n');
+  }
+
+  // Summarize
+  console.log('  요약 생성 중...');
+  const started = Date.now();
+  const rawSummary = await summarize({
+    transcript,
+    meta: {
+      title: meta?.title || 'Untitled',
+      channel: meta?.channel || 'Unknown',
+      videoUrl,
+    },
+    format: summaryFormat,
+    timelineHint,
+    useCloudSttPrompt: usedCloudStt,
+  });
+  console.log(`  요약 완료 (${((Date.now() - started) / 1000).toFixed(1)}s)`);
+
+  const { genre, keywords, body: summaryBody } = parseMetaFromSummary(rawSummary);
+
+  // Study note post-processing
+  if (route.format === 'study-note') {
+    console.log('  학습노트 생성 중...');
+    const result = buildStudyNote({
+      summaryBody,
+      url: videoUrl,
+      videoId,
+      meta,
+      transcriptStrategy: route.transcriptStrategy,
+      backendFilePath: null,
+    });
+    if (result.filePath) {
+      console.log(`  학습노트 저장: ${result.filePath}`);
+    }
+    if (flags.print) {
+      console.log('\n---BEGIN_YOUTUBE_MASTER_MARKDOWN---\n');
+      console.log(result.markdown);
+      console.log('\n---END_YOUTUBE_MASTER_MARKDOWN---\n');
+    }
+    outputSignals(route);
+    return;
+  }
+
+  // Save markdown — 도메인 스테이징 라우팅(00. Inbox/_staging/<domain>/·토글 OBSIDIAN_STAGING_BY_DOMAIN)
+  const domain = classifyDomain([genre, ...keywords, meta?.title, meta?.channel]);
+  const stagingSub = stagingEnabled() ? stagingSubdir(domain) : undefined;
+  const saveDir = flags['output-dir'] || getSaveDir(stagingSub);
+  let savedPath: string | null = null;
+  if (saveDir && route.target !== 'markdown') {
+    savedPath = await saveMarkdown({
+      meta: meta || { title: 'Untitled', channel: 'Unknown', description: '', uploaded: '', duration: 'N/A', durationSec: 0, durationIso: '', views: 0, likes: 0 },
+      videoUrl,
+      videoId,
+      summaryBody,
+      genre,
+      keywords,
+      saveDir,
+      domain,
+      artifacts: globalArtifacts || undefined,
+      cloudStt: usedCloudStt,
+    });
+    console.log(`  저장 완료: ${savedPath}`);
+  }
+
+  if (flags.print) {
+    console.log('\n---BEGIN_YOUTUBE_MASTER_MARKDOWN---\n');
+    console.log(summaryBody);
+    console.log('\n---END_YOUTUBE_MASTER_MARKDOWN---\n');
+  }
+
+  outputSignals(route);
+}
+
+// ── Transcript acquisition (auto-fallback) ──
+
+let globalSegments: TranscriptSegment[] | null = null;
+let usedCloudStt = false;
+let globalArtifacts: ArtifactInfo | null = null;
+
+async function acquireTranscript(
+  url: string,
+  meta: VideoMeta | null,
+  route: RouteDecision,
+  videoId: string,
+): Promise<string> {
+  // Pre-supplied transcript
+  if (flags['transcript-file']) {
+    console.log(`  제공된 자막 파일 사용: ${flags['transcript-file']}`);
+    return (await readFile(flags['transcript-file'], 'utf-8')).trim();
+  }
+
+  // Force Cloud STT
+  if (route.transcriptStrategy === 'cloud-stt') {
+    return acquireCloudSttTranscript(url, meta, videoId);
+  }
+
+  // Auto: try Supadata first
+  console.log('  [1/3] Supadata 자막 조회...');
+  const supadataStart = Date.now();
+  const result = await fetchSupadataTranscript(url);
+  const supadataElapsed = ((Date.now() - supadataStart) / 1000).toFixed(1);
+  if (result.text && result.text.trim().length > 0) {
+    globalSegments = result.segments;
+    const sufficiency = judgeTranscriptSufficiency(result.text, meta?.durationSec || 0);
+    if (sufficiency.sufficient) {
+      console.log(`  [1/3] Supadata 성공 (${supadataElapsed}s, ${result.text.length.toLocaleString()}자, ${result.segments.length}세그먼트)`);
+      return result.text;
+    }
+    console.log(`  [1/3] Supadata 부족 (${supadataElapsed}s): ${sufficiency.reason}`);
+  } else {
+    console.log(`  [1/3] Supadata 실패 (${supadataElapsed}s) — 자막 없음`);
+  }
+
+  // Fallback: description or Cloud STT
+  if (meta?.description && meta.description.length > 500) {
+    const hasCloudSttKeys = !!(env('ELEVENLABS_API_KEY') || env('XI_API_KEY') || env('OPENAI_API_KEY') || env('GEMINI_API_KEY') || env('GOOGLE_API_KEY'));
+    if (hasCloudSttKeys) {
+      console.log('  [2/3] Cloud STT 폴백 결정 (Supadata 실패 + STT 키 보유)');
+      return acquireCloudSttTranscript(url, meta, videoId);
+    }
+    console.log(`  [2/3] 영상 설명으로 대체 (${meta.description.length.toLocaleString()}자, STT 키 없음)`);
+    return meta.description;
+  }
+
+  // Last resort: Cloud STT
+  console.log('  [2/3] Cloud STT 최후 시도 (설명 부족 또는 없음)');
+  return acquireCloudSttTranscript(url, meta, videoId);
+}
+
+async function acquireCloudSttTranscript(url: string, meta: VideoMeta | null, videoId: string): Promise<string> {
+  usedCloudStt = true;
+  const segmentMinutes = Number(env('WHISPER_SEGMENT_MINUTES', '25'));
+  const segmentSeconds = segmentMinutes * 60;
+  const sttStart = Date.now();
+
+  const tmpDir = await mkdtemp(join(tmpdir(), 'yt-master-'));
+  try {
+    console.log('  [3/3] Cloud STT 시작...');
+    console.log('  [3/3] 오디오 다운로드...');
+    const dlStart = Date.now();
+    const audioPath = await downloadAudio(url, tmpDir);
+    console.log(`  [3/3] 오디오 다운로드 완료 (${((Date.now() - dlStart) / 1000).toFixed(1)}s)`);
+
+    console.log('  [3/3] 오디오 분할...');
+    const chunks = await splitAudio(audioPath, tmpDir, segmentSeconds);
+
+    console.log(`  [3/3] Cloud STT 전사 (${chunks.length}청크)...`);
+    const sttTranscribeStart = Date.now();
+    const result = await transcribeChunks(chunks);
+    console.log(`  [3/3] Cloud STT 전사 완료 (${((Date.now() - sttTranscribeStart) / 1000).toFixed(1)}s, ${result.transcript.length.toLocaleString()}자, 엔진: ${result.engine})`);
+
+    // Persist artifacts if Obsidian is configured
+    const saveDir = getSaveDir();
+    if (saveDir && meta) {
+      const runDir = await prepareRunDir(saveDir, meta, videoId);
+      globalArtifacts = await persistArtifacts(
+        runDir, meta, url, audioPath, chunks,
+        result.transcript, result.chunkTexts, result.engine,
+      );
+      console.log(`  아티팩트 저장: ${runDir}`);
+    }
+
+    return result.transcript;
+  } finally {
+    await rm(tmpDir, { recursive: true, force: true }).catch(() => {});
+  }
+}
+
+// ── Output signals ──
+
+function outputSignals(route: RouteDecision) {
+  if (route.target === 'web-html') {
+    console.log('\n---SIGNAL: webDeploy=html-only---');
+  } else if (route.target === 'web-deploy') {
+    console.log('\n---SIGNAL: webDeploy=deploy---');
+  } else if (route.target === 'pdf') {
+    console.log('\n---SIGNAL: pdfRequested=true---');
+  }
+}
+
+// ── Self-test ──
+
+function runSelfTest() {
+  console.log('=== youtube-master self-test ===\n');
+
+  const cases = [
+    { text: '', expect: { format: 'cards', target: 'obsidian', ts: 'auto' } },
+    { text: '요약', expect: { format: 'cards', target: 'obsidian', ts: 'auto' } },
+    { text: '짧게 요약', expect: { format: 'brief', target: 'obsidian', ts: 'auto' } },
+    { text: '상세 분석', expect: { format: 'detailed', target: 'obsidian', ts: 'auto' } },
+    { text: '학습노트로 정리해줘', expect: { format: 'study-note', target: 'obsidian', ts: 'auto' } },
+    { text: '자막만 뽑아줘', expect: { format: 'transcript', target: 'markdown', ts: 'auto' } },
+    { text: '자막 강화해서 요약', expect: { format: 'cards', target: 'obsidian', ts: 'cloud-stt' } },
+    { text: '일레븐랩스로 전사 강화', expect: { format: 'cards', target: 'obsidian', ts: 'cloud-stt' } },
+    { text: 'cc웹', expect: { format: 'cards', target: 'web-html', ts: 'auto' } },
+    { text: 'ccv웹', expect: { format: 'cards', target: 'web-deploy', ts: 'auto' } },
+    { text: 'pdf로 저장', expect: { format: 'cards', target: 'pdf', ts: 'auto' } },
+    { text: 'markdown으로', expect: { format: 'cards', target: 'markdown', ts: 'auto' } },
+    { text: '노트 + 자막 강화', expect: { format: 'study-note', target: 'obsidian', ts: 'cloud-stt' } },
+    { text: '간단히 정리 pdf', expect: { format: 'brief', target: 'pdf', ts: 'auto' } },
+    { text: '메타데이터', expect: { format: 'metadata', target: 'markdown', ts: 'auto' } },
+  ];
+
+  let passed = 0;
+  let failed = 0;
+
+  for (const tc of cases) {
+    const result = decideRoute({ intentText: tc.text });
+    const ok = result.format === tc.expect.format
+      && result.target === tc.expect.target
+      && result.transcriptStrategy === tc.expect.ts;
+
+    if (ok) {
+      console.log(`  PASS: "${tc.text || '(empty)'}" → ${result.format}/${result.target}/${result.transcriptStrategy}`);
+      passed++;
+    } else {
+      console.log(`  FAIL: "${tc.text || '(empty)'}"`);
+      console.log(`    expect: ${tc.expect.format}/${tc.expect.target}/${tc.expect.ts}`);
+      console.log(`    got:    ${result.format}/${result.target}/${result.transcriptStrategy}`);
+      failed++;
+    }
+  }
+
+  // --only tests
+  const onlyCases = [
+    { only: 'transcript', expect: 'transcript' },
+    { only: 'metadata', expect: 'metadata' },
+  ];
+  for (const tc of onlyCases) {
+    const result = decideRoute({ intentText: '', only: tc.only });
+    if (result.format === tc.expect) {
+      console.log(`  PASS: --only ${tc.only} → ${result.format}`);
+      passed++;
+    } else {
+      console.log(`  FAIL: --only ${tc.only} → expected ${tc.expect}, got ${result.format}`);
+      failed++;
+    }
+  }
+
+  // Long video test
+  const longResult = decideRoute({ intentText: '요약', durationSec: 4000 });
+  if (longResult.transcriptStrategy === 'cloud-stt') {
+    console.log(`  PASS: 긴 영상 (4000s) → cloud-stt`);
+    passed++;
+  } else {
+    console.log(`  FAIL: 긴 영상 (4000s) → expected cloud-stt, got ${longResult.transcriptStrategy}`);
+    failed++;
+  }
+
+  console.log(`\n결과: ${passed} passed, ${failed} failed`);
+
+  // Dependency checks
+  console.log('\n--- 의존성 체크 ---');
+  const checks: Array<[string, () => boolean]> = [
+    ['YOUTUBE_API_KEY', () => !!env('YOUTUBE_API_KEY')],
+    ['SUPADATA_API_KEY', () => !!env('SUPADATA_API_KEY')],
+    ['XAI_API_KEY', () => !!env('XAI_API_KEY')],
+    ['OBSIDIAN_VAULT_ROOT', () => !!env('OBSIDIAN_VAULT_ROOT')],
+    ['OPENAI_API_KEY (선택)', () => !!env('OPENAI_API_KEY')],
+    ['ELEVENLABS_API_KEY (선택)', () => !!(env('ELEVENLABS_API_KEY') || env('XI_API_KEY'))],
+  ];
+
+  for (const [name, check] of checks) {
+    console.log(`  ${check() ? 'OK' : 'MISSING'}: ${name}`);
+  }
+
+  if (failed > 0) process.exit(1);
+}
+
+// ── Help ──
+
+function printHelp() {
+  console.log(`
+youtube-master — 통합 YouTube 처리 스킬
+
+사용법:
+  npx tsx scripts/main.ts <URL> [OPTIONS]
+
+옵션:
+  --message, -m <text>     사용자 의도 (자동 라우팅)
+  --only <sub>             부분 실행: transcript, metadata
+  --format <fmt>           출력 형식: brief, cards, detailed, study-note
+  --target <tgt>           출력 타겟: obsidian, markdown, web, pdf
+  --cloud-stt              Cloud STT 강제
+  --transcript-file <path> 기존 자막 파일 공급
+  --print                  결과 stdout 출력
+  --dry-run                라우팅 결정만 출력
+  --no-obsidian            Obsidian 저장 생략
+  --output-dir <path>      저장 디렉토리 오버라이드
+  --self-test              라우팅 테스트
+  --help, -h               도움말
+`);
+}

package/skills/youtube-master/scripts/transcribe-local.ts

@@ -0,0 +1,24 @@
+import { readdir, writeFile } from 'node:fs/promises';
+import { join } from 'node:path';
+import { initEnv } from '../src/env.js';
+import { transcribeChunks } from '../src/stt.js';
+
+initEnv();
+
+const chunksDir = process.argv[2];
+const outFile = process.argv[3];
+if (!chunksDir || !outFile) {
+  console.error('usage: tsx transcribe-local.ts <chunksDir> <outFile>');
+  process.exit(1);
+}
+
+const files = (await readdir(chunksDir))
+  .filter(f => f.startsWith('chunk_') && f.endsWith('.mp3'))
+  .sort()
+  .map(f => join(chunksDir, f));
+
+console.log(`Transcribing ${files.length} chunks...`);
+const result = await transcribeChunks(files);
+await writeFile(outFile, result.transcript, 'utf-8');
+console.log(`\nDONE. engine=${result.engine}, chars=${result.transcript.length}`);
+console.log(`Saved to ${outFile}`);

package/skills/youtube-master/src/audio.ts

@@ -0,0 +1,63 @@
+import { execFile as execFileCb } from 'node:child_process';
+import { readdir } from 'node:fs/promises';
+import { join } from 'node:path';
+import { promisify } from 'node:util';
+import { which } from './util.js';
+
+const execFile = promisify(execFileCb);
+
+// YouTube는 미디어 URL의 n(스로틀) 파라미터를 플레이어 JS로 해독해야 정상 대역폭을 연다.
+// yt-dlp가 이 해독기를 EJS(외부 JS) 원격 컴포넌트로 분리하면서 기본값으로는 내려받지 않게 됐고,
+// 해독에 실패하면 약 100KB만 내려오다 HTTP 403이 난다.
+//   - --remote-components ejs:github : 해독 스크립트 다운로드 허용 (deno/node 런타임 필요)
+//   - player_client                  : 기본 android_vr은 서명 해독을 타지 않아 계속 403.
+//                                      web_embedded가 DASH 오디오를 정상 제공.
+// 참고: https://github.com/yt-dlp/yt-dlp/wiki/EJS
+const YTDLP_UNTHROTTLE_ARGS = [
+  '--remote-components', 'ejs:github',
+  '--extractor-args', 'youtube:player_client=web_embedded,web_safari,mweb',
+];
+
+export async function downloadAudio(url: string, outdir: string): Promise<string> {
+  which('yt-dlp');
+  which('ffmpeg');
+  const template = join(outdir, 'audio.%(ext)s');
+  const baseArgs = [
+    '-f', 'bestaudio[protocol^=http]/bestaudio/best',
+    '-x', '--audio-format', 'mp3',
+    '-o', template,
+  ];
+  const run = (args: string[]) =>
+    execFile('yt-dlp', args, { maxBuffer: 10 * 1024 * 1024 });
+
+  try {
+    await run([...baseArgs, ...YTDLP_UNTHROTTLE_ARGS, url]);
+  } catch (err) {
+    // 구버전 yt-dlp는 위 옵션을 모른다 — 기존 인자로 한 번 더 시도한다.
+    console.warn(`    ⚠ yt-dlp 언스로틀 경로 실패, 기본 인자로 재시도: ${(err as Error).message.split('\n')[0]}`);
+    await run(['-f', 'bestaudio/best', '-x', '--audio-format', 'mp3', '-o', template, url]);
+  }
+
+  const files = (await readdir(outdir)).filter(f => f.startsWith('audio.')).sort();
+  if (!files.length) throw new Error('오디오 다운로드 실패');
+  return join(outdir, files[0]);
+}
+
+export async function splitAudio(audioPath: string, outdir: string, segmentSeconds: number): Promise<string[]> {
+  which('ffmpeg');
+  const chunksDir = join(outdir, 'chunks');
+  const { mkdir } = await import('node:fs/promises');
+  await mkdir(chunksDir, { recursive: true });
+  const outPattern = join(chunksDir, 'chunk_%03d.mp3');
+  await execFile('ffmpeg', [
+    '-hide_banner', '-loglevel', 'error',
+    '-i', audioPath,
+    '-f', 'segment', '-segment_time', String(segmentSeconds),
+    '-c', 'copy', outPattern,
+  ]);
+  const files = (await readdir(chunksDir))
+    .filter(f => f.startsWith('chunk_') && f.endsWith('.mp3'))
+    .sort();
+  if (!files.length) throw new Error('오디오 분할 실패');
+  return files.map(f => join(chunksDir, f));
+}

package/skills/youtube-master/src/domain-staging.ts

@@ -0,0 +1,50 @@
+/**
+ * 인박스 도메인 분류 라우터 — RFC-multicontent-publishing-feed-board-2026-07-23 택소노미 사본.
+ * intake 결과를 00. Inbox/ 아래 번호 도메인 폴더로 분류 저장(소스 아닌 domain 기준). 3스킬 동기 사본
+ * (youtube-master·omni-crawl·omni-digest). 토글 OBSIDIAN_STAGING_BY_DOMAIN=0 이면 legacy 경로 유지.
+ *   폴더: 01. Daily Notes(유지) · 02. Agentic · 03. AI · 04. Stocks · 05. Tech · 06. Growth · 07. Other · 98. Archiving(구 소스폴더)
+ *   domain 6종 slug: agentic·ai·stocks·tech·growth·other (frontmatter category 진실원·여기선 키워드 폴백 분류).
+ */
+const DOMAIN_KEYWORDS: Record<string, string[]> = {
+  agentic: ['claude', 'codex', 'mcp', 'openclaw', 'subagent', 'cursor', 'vibecoding', 'vibe coding', 'coding', '코딩', '에이전트', 'agentic', 'agent tool', 'cli', 'terminal', '로컬 llm', 'local llm', '프롬프트', 'prompt', 'n8n', '자동화 에이전트', 'claude code', 'codex cli', 'anthropic sdk'],
+  stocks: ['주식', '투자', '종목', 'stock', '반도체', 'hbm', '하이닉스', '삼성전자', 'nvidia', '엔비디아', 'tsmc', 'etf', '수급', 'ipo', '매크로', '금리', '환율', '배당', '실적', '증시', '코스피', '나스닥', 's&p', '밸류에이션', '실적발표'],
+  ai: ['anthropic', 'openai', 'gemini', 'gpt', '인공지능', '생성형', 'ai 모델', 'llm 모델', '머신러닝', '딥러닝', 'ai 쇼핑', 'ai 산업', 'ai 연구', 'kimi', 'llama', 'mistral', 'ai tool'],
+  tech: ['하드웨어', '스페이스x', 'spacex', 'cpu', 'gpu', '칩셋', '반도체 산업', '맥북', 'macbook', 'mac studio', 'm5 max', 'm4 max', '디바이스', '테크 산업'],
+  growth: ['마인드셋', '자기계발', '부의', '리더십', '생산성', '비즈니스', 'business', 'leadership', 'productivity', '습관', '경영', '투자철학', '부자', '멘탈'],
+};
+
+/** domain slug → 인박스 번호 폴더명 (01. Daily Notes 는 유지·도메인은 02~07). */
+const DOMAIN_DIR: Record<string, string> = {
+  agentic: '02. Agentic',
+  ai: '03. AI',
+  stocks: '04. Stocks',
+  tech: '05. Tech',
+  growth: '06. Growth',
+  other: '07. Other',
+};
+
+/** genre·keywords·title·본문 등 신호를 합쳐 가장 매칭 많은 domain 선택(무매칭=other). */
+export function classifyDomain(signals: (string | undefined | null)[]): string {
+  const hay = signals.filter(Boolean).join(' ').toLowerCase();
+  let best = 'other';
+  let bestScore = 0;
+  for (const [domain, kws] of Object.entries(DOMAIN_KEYWORDS)) {
+    let score = 0;
+    for (const k of kws) if (hay.includes(k.toLowerCase())) score++;
+    if (score > bestScore) {
+      bestScore = score;
+      best = domain;
+    }
+  }
+  return best;
+}
+
+/** OBSIDIAN_STAGING_BY_DOMAIN=0 이 아니면 도메인 분류 저장 활성(기본 ON). */
+export function stagingEnabled(): boolean {
+  return (process.env.OBSIDIAN_STAGING_BY_DOMAIN ?? '1') !== '0';
+}
+
+/** 도메인 저장 서브디렉토리(볼트 루트 상대) — 00. Inbox/0N. <Domain>. */
+export function stagingSubdir(domain: string): string {
+  return `00. Inbox/${DOMAIN_DIR[domain] ?? DOMAIN_DIR.other}`;
+}

package/skills/youtube-master/src/env.ts

@@ -0,0 +1,36 @@
+import { readFileSync, existsSync } from 'node:fs';
+import { resolve, dirname } from 'node:path';
+import { fileURLToPath } from 'node:url';
+
+const __dirname = dirname(fileURLToPath(import.meta.url));
+export const SKILL_DIR = resolve(__dirname, '..');
+
+export function loadEnvFile(filePath: string): void {
+  if (!existsSync(filePath)) return;
+  const content = readFileSync(filePath, 'utf-8');
+  for (const raw of content.split('\n')) {
+    const line = raw.trim();
+    if (!line || line.startsWith('#') || !line.includes('=')) continue;
+    const idx = line.indexOf('=');
+    const key = line.slice(0, idx).trim();
+    const val = line.slice(idx + 1).trim().replace(/^["']|["']$/g, '');
+    if (key && !(key in process.env)) {
+      process.env[key] = val;
+    }
+  }
+}
+
+export function requireEnv(key: string): string {
+  const val = process.env[key];
+  if (!val) throw new Error(`필수 환경변수 누락: ${key}`);
+  return val;
+}
+
+export function env(key: string, fallback = ''): string {
+  return process.env[key] || fallback;
+}
+
+/** Load the single .env from skill root */
+export function initEnv(): void {
+  loadEnvFile(resolve(SKILL_DIR, '.env'));
+}