naver-news-briefing

Search, brief, and monitor 대한민국 Naver News via the Naver Search API using natural-language Korean queries. Use when the user wants 네이버 뉴스 브리핑, 최근 N일 뉴스 요약, 제외어 포함 뉴스 검색, 여러 질의를 한 번에 묶은 멀티 브리핑, 키워드 그룹 저장/관리, 특정 키워드의 지속 감시 규칙 추가/목록/삭제, 채팅형 자동화 요청을 watch/group 설정으로 바꾸기, or cron-friendly operator guidance for periodic news checks. Prefer this skill for Korean Naver-news workflows backed by local persistent state.

Use the CLI script at scripts/naver_news_briefing.py.

Onboarding

• Treat Naver Search API credentials as mandatory before the first real use.
• Tell the user early that search/briefing/watch flows will fail until client_id and client_secret are stored.
• If the user has not completed setup yet, direct them to run:
  • python scripts/naver_news_briefing.py setup --client-id ... --client-secret ...
  • python scripts/naver_news_briefing.py check-credentials --json
• Present setup as the first-run path, not an optional advanced step.
• When helping with installation or first use, mention that credentials are stored in data/config.json and use DPAPI-backed secret storage on Windows when possible.

Workflow

1. Store credentials once before any search/brief/watch command.
   • python scripts/naver_news_briefing.py setup --client-id ... --client-secret ...
   • Verify with python scripts/naver_news_briefing.py check-credentials --json
2. Run a one-shot briefing.
   • python scripts/naver_news_briefing.py search "최근 3일 반도체 뉴스 브리핑 -광고"
   • Add --json for machine-readable output.
3. Manage persistent watch rules with optional operator metadata.
   • Add: python scripts/naver_news_briefing.py watch-add semiconductor "최근 7일 반도체 -광고" --label "반도체 감시" --tag watch --template watch-alert
   • List: python scripts/naver_news_briefing.py watch-list
   • Remove: python scripts/naver_news_briefing.py watch-remove semiconductor
   • Check: python scripts/naver_news_briefing.py watch-check semiconductor --json
4. Manage persistent keyword groups for recurring briefings.
   • Add: python scripts/naver_news_briefing.py group-add market-watch "최근 3일 반도체 -광고" "오늘 AI 데이터센터 -주가" --label "아침 시장" --tag 테크 --context "오전 보고용" --template morning-briefing
   • List: python scripts/naver_news_briefing.py group-list
   • Inspect one group: python scripts/naver_news_briefing.py group-list market-watch --json
   • Update: python scripts/naver_news_briefing.py group-update market-watch --add-query "배터리 공급망 -광고" --tag 공급망 --template analyst
   • Remove: python scripts/naver_news_briefing.py group-remove market-watch
5. Run combined briefings.
   • python scripts/naver_news_briefing.py brief-multi --group market-watch
   • python scripts/naver_news_briefing.py brief-multi --group market-watch --query "환율 뉴스" --template morning-briefing --json
   • If --template is omitted, prefer the saved group template when present.
6. Convert chat-style automation requests into structured plans.
   • Inspect plan: python scripts/naver_news_briefing.py plan "반도체 뉴스 1시간마다 모니터링해줘" --json
   • Build an OpenClaw-friendly integration bundle: python scripts/naver_news_briefing.py integration-plan "반도체 뉴스 1시간마다 모니터링해줘" --json
   • Support practical Korean patterns such as 매일 아침 7시에 반도체랑 AI 데이터센터 뉴스 브리핑해줘 and 증권사 리포트 빼고 삼성전자 뉴스 계속 체크해줘.
7. Materialize plans into persistent configs.
   • Watch: python scripts/naver_news_briefing.py plan-save "반도체 뉴스 1시간마다 모니터링해줘" --as watch --name semiconductor-hourly
   • Group: python scripts/naver_news_briefing.py plan-save "매일 아침 7시에 반도체랑 AI 데이터센터 뉴스 브리핑해줘" --as group --name morning-tech --label "아침 브리핑"

Behavior

• Parse positive keywords and -제외어 using the upstream tab-search policy.
• Interpret recent-news phrases such as 오늘, 최근 3일, 최근 2주, 한달, 이번주, 지난주 as a date window and remove that phrase from the API search query.
• Normalize more natural Korean sentence inputs by stripping request phrases, common 조사, duplicate tokens, and Korean exclusion phrases such as A 말고, B 빼고, C 제외 into -제외어 tokens.
• Parse practical schedule intent from Korean chat requests into interval, daily, weekly, or manual plans.
• Preserve operator-facing metadata on saved watch/group entries: label, tags, template, schedule, operator_hints, and original request context.
• Use DPAPI-backed secret storage on Windows when possible.
• Deduplicate watch notifications by (watch_id, link) so repeated cron runs emit only newly seen items.
• plan returns cron-friendly operator hints, recommended commands, and a storage-target recommendation (watch vs group).
• integration-plan returns a more practical operator bundle: save command, run command, schedule object, cron line, OpenClaw-friendly systemEvent text, and a Korean confirmation summary.
• plan-save materializes a parsed plan into a saved watch or group configuration without owning cron wiring itself.
• brief-multi returns chat-friendly combined text by default and structured JSON with --json.

Notes

• Read references/upstream-notes.md before major edits.
• The skill uses headline/summary metadata from the Naver Search API. It does not fetch or summarize full article bodies.
• Keep additions additive: preserve existing search, watch-add, watch-list, watch-remove, and watch-check flows.
• Public user-facing documentation lives in README.md; keep it Korean-first and concrete.