news-to-note

Intelligent News Interpretation — Understand every piece of news with your notes knowledge base

Core Philosophy

News alone has no value; news × your knowledge base = valuable insight.

Any app can search news, but only WPS Notes knows what the user cares about, what projects they are working on, what meetings they have attended, and what ideas they have recorded. The core capabilities of this Skill:

Save news/information into WPS Notes (precipitate as knowledge assets)
In the user'sentire note libraryretrieve related content (work notes, meeting records, project documents, etc.)
Connect news with the user's existing knowledge to producepersonalized insight

Two Workflows

Workflow	Trigger Scenario	Output
A — News Collection	"Find news" "hot topic summary" "news briefing"	Daily briefing note
B — News Interpretation	The user encounters any specific news/information	Note archive + personalized in-depth analysis

Decision rules:

The user encountersspecific news(sends a link, mentions a topic, pastes the body text, discusses a news event) → B
The user requestsbatch search, or a briefing summary → A
When ambiguous, default to B

Not applicable to: regular reading, writing, and editing of notes (use the wps-note Skill), saving non-news content, and pure knowledge Q&A.

Workflow A: News Collection

Batch search → deduplicate and filter → write into briefing note.

Steps

Construct search terms: include at least topic words + time-sensitive words, use site: to limit sources, 2-4 query groups. See template in query-recipes.md。
Normalize and deduplicate: remove utm_* and other tracking parameters, deduplicate by URL, avoid duplicate writes.
Locate target note: title format 新闻简报 YYYY-MM-DD, search first then create.
Write content: refresh the outline to get block_id, use edit_block(op=insert) to append <h2> title + <p><a> link + <p> summary.
Sync：sync_note, report the number of inserted items.

Workflow B: News Interpretation (Core)

This is the core differentiating capability of this Skill. Triggered when the user encounters any news.

B1. Obtain News Content

Input Type	Acquisition Method
Has URL	`web_fetch` fetch the full text; on failure fall back to `web_search`
Has topic but no link	`web_search` search → select the best 1-2 articles → `web_fetch` fetch
Directly paste body text	Use directly, extract title and source

Clean content: remove non-body content such as navigation bars, ads, and comment sections.

B2. Save to WPS Notes

Use the article title as the note title. Write the title and metadata in one go,the body must be written in multiple batches。

One-time write — create note + metadata:

create_note(title=文章标题)
get_note_outline → 获取锚点 block_id
edit_block(op=insert) 写入元信息：

<h1>文章标题</h1>
<p><tag>#新闻</tag> <tag>#新闻//领域标签</tag></p>
<highlightBlock emoji="🔗" highlightBlockBackgroundColor="#E6EEFA" highlightBlockBorderColor="#98C1FF">
  <p><strong>来源：</strong><a href="原文URL">来源名称</a></p>
  <p><strong>发布时间：</strong>YYYY-MM-DD</p>
</highlightBlock>
<h2>正文</h2>

Append in batches — write the body in segments:

Write the body in batches by paragraph, using the value returned by the previous insert each time as the last_block_id anchor:

edit_block(op=insert, anchor_id=last_block_id, position=after)
  → 写入 3-5 个 <p> 段落
  → 拿到新的 last_block_id

edit_block(op=insert, anchor_id=新的last_block_id, position=after)
  → 写入下一批 3-5 个 <p> 段落
  → 重复直到正文写完

Key rules:

Metadata such as title, tags, and source can be written in one go.
The body must not be written in one go, it must be appended in batches, 3-5 <p> paragraphs per batch.
For consecutive inserts, directly use the returned last_block_id as the anchor, no need to refresh the outline each time.
After the last batch is written, call sync_note to sync.

Tag strategy: always add #新闻, and add secondary tags based on content such as #新闻//AI、#新闻//金融, 1-3 per article.

B3. Search the user's knowledge base (key step)

Core differentiator: don't just search news, search all of the user's notes.

Extract 3-5 keywords and search acrossthe entire note libraryin multiple dimensions:

Full-text search by keyword (company names, technology names, industry names, people's names)
Search historical news by domain tags
Search recent notes by time range

Work notes, meeting records, and project documents are all valuable related sources. For detailed search strategies, see knowledge-search-strategy.md。

B4. Read related notes

get_note_outline preview the structure and judge relevance
For highly relevant notes, use read_note / read_section to read, up to 5 articles
Focus on extracting the user's personal views, decisions, and plans

B5. Produce personalized insights

It is not a news summary, but "what this news means to me."

Answer structure:

News highlights — summarize in 2-3 sentences
Relevance to you — cite the user's notes: "You mentioned in [note title]...", and point out the potential impact on current work
Deep insights — trend judgment, comparative analysis, action recommendations
Further reading — recommend related notes in the note library worth rereading

Requirement: if removing the user's note associations makes the answer mediocre, that is what shows the insight was done right. If there are no related notes, state so honestly and provide a general analysis.

Examples

Example 1: The user sends a link

The user says: "Help me look at this https://36kr.com/p/xxx news about OpenAI funding"

Actions:

web_fetch fetch the full text → extract the title "OpenAI receives 110 billion USD in funding"
create_note → edit_block write metadata and tags → append body paragraphs in batches → sync_note
search_notes(keyword="OpenAI") + search_notes(keyword="AI 融资") + search_notes(keyword="大模型") search the knowledge base
Find that the user has an "AI Strategy Planning" note and a "Competitor Analysis" document → read_note read
Output: news highlights + "You mentioned in 'AI Strategy Planning' that you are watching OpenAI's API pricing strategy; this funding may accelerate the pace of its price cuts" + action recommendations

Result: the news has been saved to notes, and the user receives personalized in-depth analysis.

Example 2: User chats about a news topic

User says: "How has DeepSeek been doing lately?"

Action:

web_search("DeepSeek 最新动态 2026") → Find the most authoritative report → web_fetch fetch
Save to notes, tag with #新闻//AI tags
search_notes(keyword="DeepSeek") + search_notes(keyword="开源模型") Search the knowledge base
Discover that the user previously saved a news article titled "DeepSeek-V3 Release" → Compare timelines
Output: Latest updates + "Comparing with the DeepSeek-V3 report you previously saved..." + Trend analysis

Result: The news has been saved to notes, and the user sees the development trajectory of the event.

Example 3: Batch search briefing

User says: "Help me search for today's AI news"