chinese-news-digest

name	chinese-news-digest
description	Fetch trending content from 3 Chinese news aggregation platforms (NewsNow, TopHub Today Tech, SoPilot Hot Tweets), extract top news items, categorize each for the appropriate Slack channel, perform web research, and post x-to-slack format 3-message threaded summaries to Slack. ALWAYS invoke when the user asks to "run Chinese news digest", "중국 뉴스 다이제스트", "chinese-news-digest", "중국 뉴스 정리", "중국 트렌딩", "NewsNow 뉴스", "今日热榜", "SoPilot", "중국 뉴스 슬랙에 올려줘", "중국 뉴스 수집", "chinese news to slack", "중국 핫토픽", or wants daily Chinese-language tech/social trend intelligence posted to Slack. Do NOT use for single tweet analysis (use x-to-slack). Do NOT use for Bespin news email processing (use bespin-news-digest). Do NOT use for general web research without Chinese source intent (use parallel-web-search). Do NOT use for HuggingFace trending scan (use hf-trending-intelligence).
metadata	{"author":"thaki","version":"2.1.0","category":"pipeline"}

Chinese News Digest Pipeline

Fetch trending topics from 3 Chinese aggregation platforms, select the top 5 most newsworthy items, categorize each for the right Slack channel, run web research per item, and post x-to-slack format 3-message threads.

Pattern: mirrors bespin-news-digest -- sequential processing with mandatory WebSearch per item. Direct scripts/slack_post_message.py for Slack posting (not MCP slack_send_message).

Sources

#	Platform	URL	Content Type
1	NewsNow (多平台聚合)	https://newsnow.busiyi.world	Zhihu, Weibo, Bilibili, Hupu, V2EX aggregation
2	TopHub Today (今日热榜 - Tech)	https://tophub.today/c/tech	GitHub Trending, Product Hunt, Hacker News, tech circle
3	SoPilot (X Hot Tweets)	https://sopilot.net/zh/hot-tweets	X/Twitter viral posts, opinion trends

Slack Channel Registry

Categorize each news item to the most appropriate channel:

Channel Name	Channel ID	When to Use
deep-research-trending	C0AN34G4QHK	AI/ML research, deep tech, academic papers, benchmarks
ai-coding-radar	C0A7K3TBPK7	AI coding agents, dev tools, Claude/Cursor/Copilot news
press	C0A7NCP33LG	General tech news, product launches, business/media, social trends
효정-할일	C0AA8NT4T8T	Actionable items requiring personal follow-up
효정-insight	C0A8SSPC9RU	Strategic insights, market analysis

Default channel: #press (C0A7NCP33LG) when categorization is ambiguous.

Constraints

Freedom level: Low -- follow the 6-step workflow exactly. Do not skip steps, reorder phases, or invent new output formats.

• All Slack messages in Korean
• Use scripts/slack_post_message.py for posting (user identity, not bot)
• NEVER use slack_send_message MCP tool for text posts
• Message 1 is plain text only (no mrkdwn formatting)
• Messages 2-3 use mrkdwn formatting (*bold*, _italic_)
• Sequential processing: post all 3 messages for item N before starting item N+1
• Rate limiting: 2-second pause between items to avoid Slack throttling
• Do not add commentary, emojis, or embellishments beyond the specified format
• Match output length to content -- do not pad thin news items with speculation

Workflow

Step 1: Fetch Content from All 3 Sources

Use WebFetch to retrieve content from each source URL. If a source times out, retry once. If it still fails, skip that source and proceed with remaining sources (log a warning).

Step 2: Extract Top News Items

From the combined content across all 3 sources, identify the top 5 most newsworthy items using these criteria:

1. Impact: How broadly does this affect the tech/AI/product ecosystem?
2. Novelty: Is this genuinely new information (not rehashed)?
3. Relevance: Does it matter for AI platform strategy, GPU cloud, or dev tools?
4. Engagement: High social engagement signals (likes, comments, shares)
5. Diversity: Avoid clustering -- pick items from different topics/sources

For each item, extract:

• Title: Original Chinese title (or closest equivalent)
• Summary: 2-3 sentence summary in Korean
• Source platform: Which of the 3 sources it came from
• Original URL: Direct link if available; otherwise the source platform URL
• Category: Channel assignment (Step 3)

Step 3: Categorize for Slack Channels

Assign each item to a Slack channel from the Channel Registry above. Classification rules:

Topic Pattern	Channel
LLM, foundation model, benchmark, training, inference optimization	deep-research-trending
AI agent, coding assistant, IDE, developer tool, MCP, Claude, Cursor, Copilot	ai-coding-radar
Product launch, funding, IPO, social trend, general tech, consumer AI	press
Actionable for personal workflow (tool to adopt, account to create)	효정-할일
Strategic market signal, competitive intelligence	효정-insight

Step 4: Web Research per Item

For each of the 5 items, run WebSearch with 2-3 targeted queries to gather:

• Additional context and background
• English-language coverage of the same topic
• Related developments or competing perspectives

Step 5: Post to Slack (3-Message Thread per Item)

Post each item as a 3-message thread using scripts/slack_post_message.py.

Message 1: Title + URL (initial post)

python3 scripts/slack_post_message.py \
    --channel "{channel_id}" \
    --message "{Korean headline}\n{source_url}"

Strict rules for Message 1:

• EXACTLY 2-3 lines: Korean headline + URL
• NO emojis, NO bold (*), NO italic (_), NO mrkdwn formatting
• Plain text only
• One sentence summarizing the core insight in Korean
• Source URL on the next line

Parse the JSON output to capture ts for threading.

Message 2: Detailed Summary + Research (thread reply)

python3 scripts/slack_post_message.py \
    --channel "{channel_id}" \
    --message "{detailed_message}" \
    --thread-ts "{message_1_ts}"

Format (Korean, mrkdwn):

*[Source Platform] 상세 분석*

*핵심 내용*
{3-5 bullet points summarizing the news item}

*배경 및 맥락*
{Background context from web research}

*웹 리서치 결과*
{Key findings from WebSearch -- additional coverage, expert opinions}

*관련 링크*
- {link 1}
- {link 2}

Message 3: AI/Strategic Insights (thread reply)

python3 scripts/slack_post_message.py \
    --channel "{channel_id}" \
    --message "{insights_message}" \
    --thread-ts "{message_1_ts}"

Format (Korean, mrkdwn):

*AI/전략 인사이트*

*시사점*
{What this means for AI platform strategy, GPU cloud, or the broader ecosystem}

*액션 아이템*
{Specific actions to consider -- if any}

*경쟁 영향*
{Impact on competitors or market positioning -- if relevant}

Step 6: Completion Summary

After all 5 items are posted, output a summary:

중국 뉴스 다이제스트 완료
- 처리된 항목: 5
- 채널별 분포: #press (N), #ai-coding-radar (N), #deep-research-trending (N)
- 소스별 분포: NewsNow (N), TopHub (N), SoPilot (N)

Verification Protocol

Before declaring the pipeline complete, verify:

1. All 3 sources were attempted (log skipped sources with reason)
2. Each posted item has exactly 3 messages (initial + 2 thread replies)
3. Message 1 contains NO mrkdwn formatting (plain text only)
4. Thread replies reference the correct message_1_ts
5. No duplicate stories were posted across channels

VERDICT: PASS — all 5 items posted with 3-message threads
VERDICT: FAIL — {N} items failed: {reason}

Report the verdict honestly at Step 6.

Output Discipline

• Do not add features or sections beyond the 3-message format
• If a news item has thin content, write a shorter summary -- do not pad
• Do not speculate about implications without evidence from WebSearch
• Match Korean output length to actual content density
• If fewer than 5 newsworthy items exist, post only what qualifies -- do not fill quota with low-quality items

Honest Reporting

• Report outcomes faithfully: if sources returned no usable content, say so
• Never claim "all items posted successfully" without verifying each ts response
• If WebSearch returns no relevant results for an item, state that explicitly in Message 2 rather than fabricating context
• Report source fetch failures, Slack errors, and skipped items transparently

Required Tools

Tool	Purpose
WebFetch	Fetch content from 3 source URLs
WebSearch	Research each news item for context
Shell	Execute scripts/slack_post_message.py for Slack posting

Gotchas

Symptom	Root Cause	Correct Approach
WebFetch returns garbled HTML with no readable content	Source uses heavy JS rendering	Try fetching the mobile/API variant, or extract what text is available and supplement with WebSearch
WebFetch timeout on any source	Network instability or source downtime	Retry once, then skip source with warning and proceed with remaining sources
Same story appears in all 3 sources	Viral content cross-platform	Consolidate into one item; note multi-source coverage in Message 2
slack_post_message.py returns {"ok": false}	Channel ID mismatch or token expired	Verify channel ID from registry; check SLACK_USER_TOKEN in .env
Thread replies post as top-level messages	Missing or malformed --thread-ts	Ensure ts is parsed from Message 1 JSON output as string (not truncated float)
Chinese content renders as mojibake in Slack	Encoding issue in shell escaping	Use heredoc or write message to temp file, then pass via --message-file if available
Fewer than 5 newsworthy items across all sources	Low-activity period or source issues	Post only qualifying items; do not pad with low-quality fillers

Example Invocation

/chinese-news-digest
chinese-news-digest
중국 뉴스 다이제스트 실행
중국 뉴스 정리해줘
중국 트렌딩 슬랙에 올려줘