trend-report

trendradar

When to activate

• Chinese: “produce a daily report” / “weekly report” / “monthly report” / “trend digest” / “compare with last week” / “compare with last month” / “a brief for the boss” / “today’s summary” / “this week’s summary” / “this month’s summary” / “a report for the past 7 days”
• English:

  summary report

  /

  digest

  /

  daily/weekly/monthly digest

  /

  period over period

trend-search

trending-now

Workflow

1. Resolve the time range (required) Call

   mcp__trendradar__resolve_date_range

   to convert natural-language expressions such as

   today / yesterday / this week / last week / this month / last month / Q1 / the past 7 days / April 1 to April 15

   into ISO

   start_date

   /

   end_date

   values. Cross-year, cross-month, and cross-quarter boundaries must always go through this step; do not calculate them manually.
2. Try the one-stop interface first If the user wants a “standard daily/weekly/monthly report,” try the following first:

   mcp__trendradar__generate_summary_report(
     date_range={"start_date": ..., "end_date": ...},
     report_type="daily" | "weekly" | "monthly"
   )

   If the returned structure already covers the user’s requirements, proceed directly to step 5 (apply a template).
3. Combine manually (fallback) If the output from

   generate_summary_report

   is insufficient—for example, the user wants an industry breakdown or a trend-jacking opportunity score—combine the following calls:
   • Daily data: Loop through every date in the range and call

     mcp__trendradar__get_news_by_date(date=...)

   • Cross-period insights: Use

     mcp__trendradar__analyze_data_insights(date_range=..., dimensions=[...])

     for topic trends, platform distribution, and sentiment
   • High-engagement topics: Use

     mcp__trendradar__get_trending_topics(date_range=..., limit=10)

4. Handle comparison periods When the user includes

   comparison / vs / period over period / year over year / versus last week / compared with last month

   :

   mcp__trendradar__compare_periods(
     period_a={"start_date": ..., "end_date": ...},  // 当期
     period_b={"start_date": ..., "end_date": ...}   // 对照期
   )

   period_a

   is always the “primary period” explicitly mentioned by the user, and

   period_b

   is the reference period being compared against it.
5. Select the output template (choose exactly one; do not mix them) Match the template based on the user’s audience cues. If there are no cues, default to Executive Summary.

   • boss / CEO / founder / management / executives / board / one-sentence summary

     → Executive Summary

   • media / content industry / platform / IP / policy / industry insights

     → Media Industry

   • ad placement / advertising / trend-jacking / marketing / brand / reach restrictions / sensitive terms / traffic

     → Campaign Team
6. Append a chart-friendly data block (required) The end of the report must include a Markdown table that can be pasted directly into Feishu Docs / Notion / Lark Sheets, along with a one-sentence “how to chart” hint (bar chart / line chart / pie chart).

Output Templates

Template A — Executive Summary (C-suite, default)

Keep the entire report under 400 words. Use a restrained tone and lead with the conclusions.

# {Period title, e.g., Weekly Report for 2026-04-22 ~ 2026-04-28}

## TL;DR
{Up to 3 lines, with one core insight per line. Start with the change, then the magnitude.}

## Key Events
- {Event 1: what happened + magnitude}
- {Event 2: what happened + magnitude}
- {Event 3: what happened + magnitude}

## Key Metrics
| Metric | Current Period | Previous Period | Change |
| --- | --- | --- | --- |
| Total topic count across all platforms | … | … | ±x% |
| Peak engagement of the leading topic | … | … | ±x% |
| Number of high-emotion-intensity events | … | … | ±x% |

## What We Should Do
1. {Action 1: actionable, with an implied owner}
2. {Action 2}
3. {Action 3}

Template B — Media Industry

Five fixed sections, each containing 3–5 bullet points, with every bullet ending in a one-line interpretation.

# {Period title} · Media Industry Brief

## 1. Major Industry Events
- {Event} — {1-line interpretation}

## 2. Platform Updates
- {Platform/product/rule change} — {1-line interpretation}

## 3. Content Format Trends
- {Content format/subject/format} — {1-line interpretation}

## 4. Leading IP Changes
- {IP/show/creator} — {1-line interpretation}

## 5. Policy Direction
- {Policy/regulation/platform rule} — {1-line interpretation}

Template C — Campaign Team

# {Period title} · Trend-Jacking Brief

## Top 10 Topics by Engagement This Period
| # | Topic | Main Platform | Peak Engagement | Relevant Brands | Trend-Jacking Opportunity Score (1–5) |
| - | --- | --- | --- | --- | --- |
| 1 | … | … | … | … | … |

> Scoring criteria: 5 = currently viral + strong brand relevance; 3 = trending but requires creative adaptation; 1 = high risk or declining.

## Risk Terms / Reach-Restricted Topics
- {Term or topic} — {Risk type: reach restriction / sensitive / policy / brand conflict}

## Top 3 Trend-Jacking Opportunities for Next Week
1. {Topic} — {Suggested angle, 1 line}
2. {Topic} — {Suggested angle, 1 line}
3. {Topic} — {Suggested angle, 1 line}

Chart-Friendly Data Block (required for all templates)

## Appendix: Paste-Ready Data Block
| Dimension | Value |
| --- | --- |
| … | … |

> How to chart: We recommend using a “{bar chart / line chart / pie chart}” to display {field}.

• Time-series trends → line chart
• Platform/category comparisons → bar chart
• Share and distribution → pie chart

Edge Cases

• Insufficient data: When

  get_news_by_date

  or

  generate_summary_report

  returns empty results for certain dates, first list the missing dates, then recommend that the user run

  mcp__trendradar__trigger_crawl(date=...)

  to retrieve the missing data, and finally produce a reduced report based on the available data with the label “Coverage X/Y days.” Do not treat days with missing data as 0.
• Cross-year / cross-month / cross-quarter: Always call

  resolve_date_range

  first; manually concatenating ISO dates is prohibited. Use complete date ranges in cross-year report titles, such as

  2025-12-29 ~ 2026-01-04

  , to avoid ambiguity.
• Unequal comparison-period lengths:

  period_a

  and

  period_b

  must contain the same number of days. If their lengths differ, extend the shorter period or truncate the longer period as appropriate to align them, and note “Comparison period aligned to N days” in the TL;DR.
• MCP errors / timeouts: Retry once first. If the retry still fails, fall back to a combination of

  get_latest_news + analyze_data_insights

  , and add the header “Data source: live crawl (not a complete archive)” to the report.

Don’ts

• Do not combine “Executive Summary + Media Industry + Campaign Team” in one report. Each report serves one audience—choose exactly one of the three.
• Do not insert URLs into the main body unless the user explicitly requests “include the original links.” When needed, add a separate

  ## Original Links

  appendix.
• Do not fabricate numbers that the MCP did not return. Do not use vague qualifiers such as “approximately / estimated / roughly.” If data is missing, explicitly mark it as

  N/A

  or explain it under Edge Cases.
• Do not write action recommendations as empty phrases such as “increase attention” or “continue monitoring.” Either specify a concrete topic, concrete action, and concrete time window, or omit the section.
• Do not skip

  resolve_date_range

  and manually construct ISO dates.
• Do not mix daily and weekly reporting scopes within the same report, such as placing “today’s peak” and “weekly total” in the same table without labels.

See also

• ../docs/mcp-tools.md

  — Complete field descriptions and return structures for the trendradar MCP tools.