Kr naver insight

Korean Naver Blog Insights

Given a company and a list of Naver blogger IDs, read their recent posts, extract the body text, and produce a structured Markdown digest that plugs into the Street / Alternative Views section of a full memo.

When To Use

• After kr-naver-blogger has produced a dated JSON file for a ticker.
• When the user supplies an explicit list of blogger IDs to read.
• When a memo needs source type 4 material (independent long-form analysis) and Naver is the right venue for it.

Do not use this skill to:

• Produce a final thesis or stance — that belongs in kr-stock-analysis.
• Cite bloggers as primary sources for material facts. Treat them as secondary perspectives.

Pipeline

1. Fetch — scripts/fetch-blog-posts.js reads the bloggers list, walks each blogger's PostList, filters posts that mention the company name or ticker, and pulls the body of each relevant post through kr-naver-browse. Posts are cached per logNo under the cache dir; a post is immutable, so cache never expires.
2. Summarize — scripts/summarize-insights.js reads the posts JSON and renders a Markdown digest grouped by blogger, with per-post date, title, snippet, and a source URL.

Output Contract

The summarize step writes a Markdown file that looks like:

# Naver Blog Insights: 엘앤에프 (066970)

기준일: 2026-04-12

## Blogger Coverage Summary

| Blogger | Posts | Coverage Window | Latest |
|---------|-------|-----------------|--------|
| mooyoung_2022 | 5 | 2025-10-12 ~ 2026-04-12 | 2026-04-12 |

## mooyoung_2022 — Recent Posts

### 실리를 찾는 국가 지도자 (2026-04-12)

> 에너지 안보 앞에 다른 선택은 없다. 배터리 양극재를 만드는 기업인 엘앤에프에…

📎 https://blog.naver.com/PostView.naver?blogId=mooyoung_2022&logNo=224249188355

Reference: references/output-format.md.

Operating Rules

1. Never paraphrase into a thesis. The digest quotes the post and links the source. Synthesis belongs in the memo itself.
2. Quote, don't invent. Snippets must be lifted verbatim from the fetched text, with ellipsis for truncation. No reworded summaries.
3. Always include the source URL for each post. A memo writer needs the link to cite.
4. Truncate each post body to 3000 chars during fetch (handled by kr-naver-browse). Truncate snippets in the digest to 500 chars.
5. Skip posts that do not mention the company or ticker in title or first 500 characters of body.
6. Stay per-blogger deterministic — sort posts by date descending, fall back to logNo descending.
7. Respect the cache — same logNo should never be re-fetched unless --no-cache is set.

Invocation

# Step 1: fetch (either from blogger list JSON or explicit IDs)
node skills/kr-naver-insight/scripts/fetch-blog-posts.js \
  --input .tmp/naver-blog-cache/bloggers/066970/2026-04-12.json \
  --company "엘앤에프" --max-posts 5 --output /tmp/posts.json

# or
node skills/kr-naver-insight/scripts/fetch-blog-posts.js \
  --company "엘앤에프" --ticker 066970 \
  --bloggers mooyoung_2022 --max-posts 5 --output /tmp/posts.json

# renamed tickers can pass legacy names too
node skills/kr-naver-insight/scripts/fetch-blog-posts.js \
  --company "리가켐바이오" --aliases "레고켐바이오,리가켐바이오사이언스" \
  --ticker 141080 --bloggers foreconomy --max-posts 5 --output /tmp/posts.json

# Step 2: summarize
node skills/kr-naver-insight/scripts/summarize-insights.js \
  --input /tmp/posts.json --output /tmp/insights.md

Known Issue

• Sandbox runtime bug: this skill depends on kr-naver-browse, which starts a local gstack browse server. In some Codex sandbox environments, local 127.0.0.1 listen is blocked with EPERM, so the fetch step cannot run inside the sandbox.
• What it looks like: fetch-blog-posts.js may report browse startup failures such as No available port after 5 attempts or listen EPERM: operation not permitted.
• How to interpret it: treat this as an environment bug first, not as proof that the blogger has no relevant posts.
• Workaround: rerun fetch-blog-posts.js outside the sandbox or with elevated execution. After the recent fix, these failures should also be recorded in posts.json.meta.errors instead of being silently flattened into posts: [].