Last-30-Days Cross-Platform Trend Researcher Pro

What you get: a structured workflow that turns a stuck problem into a repeatable routine, delivering reusable artifacts and templates.

Introduction

[TRACE 5.0 positioning] This skill is a hands-on tool. It ships reusable artifacts and templates only, never vague advice. The original GitHub skill (mvanhorn/last30days-skill) turns "what happened around this topic across platforms in the last 30 days" into an explicit retrieval task instead of letting the model answer from stale memory. The Pro version upgrades it into a full research pipeline: multi-source aggregation with deduplication, sentiment and stance distribution by dimension, and a trend report with a timeline and a map of where opinions diverge, every claim carrying its source and timestamp so the output drops straight into weekly reports, competitor briefs and editorial planning.

What I can do

• Break the topic into a query matrix (core term + synonyms + English terms + long-tail) spanning news, communities, developer platforms, video and Q&A sources.
• Aggregate and deduplicate: collapse multiple retellings of one event into a single fact entry, preserving the earliest source and the follow-up chain.
• Analyze sentiment and stance in four buckets (positive / neutral / negative / contested), extracting representative quotes and the actual points of disagreement rather than one blanket label.
• Build a timeline: order key moments by date, marking the trigger event, amplification nodes and the cool-down point.
• Produce a trend report: direction of attention, driving factors, under-discussed gaps, and actionable implications for your business.
• Supply reusable retrieval and reporting templates so the next topic is plug-and-play.

What I cannot do

• I do not provide real-time data feeds or perform scraping on your behalf; conclusions depend on public sources reachable by you or the runtime.
• I do not predict stock prices, elections or match results, and I do not give investment advice.
• I do not profile or track individuals, and I do not collect personal private information.
• I do not present unverified rumors as fact; anything without cross-confirmation is labeled "single source, unverified".

Trigger phrases

last 30 days trends, trend research, topic scan, cross-platform research, industry pulse, sentiment analysis, opinion distribution, market brief, competitor moves, editorial research, attention shift, timeline review, source aggregation, monthly trend report, discourse mapping

Input specification

Provide:

1. The topic or keywords (required), e.g. "AI coding assistants", "domestic databases", "a cross-border e-commerce category".
2. Time window (defaults to the last 30 days, adjustable).
3. Lens (required, multi-select): product moves / technical progress / user complaints / competitor actions / policy and industry rules / content opportunities.
4. Intended use: weekly report, competitor brief, editorial planning, decision input.
5. Optional: keywords to exclude, platforms that must be covered, language scope (Chinese / English / both).

Output specification

Delivered as Markdown artifacts:

• Executive summary: up to five one-line conclusions.
• Fact table: date | event | source type | key detail | confidence (multi-source confirmed / single source).
• Timeline: key moments in date order with warm-up and cool-down inflection points.
• Sentiment and stance table: dimension | positive | neutral | negative | contested point | representative quote.
• Trend read: direction, drivers, risk signals, under-discussed gaps.
• Action plan: 3 to 5 concrete steps tailored to your stated use.
• Source list: title, platform, date, so anything can be re-checked.

Usage examples

Example 1

Input: "Topic: AI coding assistants; lens: user complaints + competitor actions; use: product weekly report." Output: Four summary lines, e.g. "negative feedback clusters around lost long context and out-of-scope edits, not completion accuracy"; a fact table of version releases and pricing changes across vendors in the window; a sentiment table showing how paid and free users differ in what they care about; a trend read stating the competition is shifting from completion quality to edit controllability; and an action item to add change-scope visualization to the backlog at high priority.

Example 2

Input: "Topic: an open-source database community; window: last 45 days; lens: technical progress + policy and industry rules; language: both." Output: A bilingual query matrix; a timeline marking three notable releases and one license change; a divergence table laying out the two camps in the license debate with their respective arguments; a risk signal noting that several downstream distributions have not stated a position, so fork discussions may follow; and a source list grouped by platform, flagging which conclusions are multi-source confirmed.

Example 3

Input: "Topic: home energy storage; use: editorial planning; find angles nobody has covered well." Output: A list of high-attention angles with a crowding score (dense but highly repetitive topics called out separately); a gap list, e.g. "post-installation service and warranty fulfillment experience" and "real degradation feedback in humid southern regions" lack first-hand coverage; each gap paired with a suggested format, the kind of interviewee to seek, and the facts that must be verified; plus five ready-to-start headline drafts with their verification checklists.

Error handling

• Topic too broad: return three narrowing options (by audience / by scenario / by time slice) and wait for your pick before running.
• Nothing recent found: state plainly that no public information within the window was reachable, then offer alternatives (widen the window, change the query terms, pivot to an adjacent topic). Never fabricate entries.
• Sources contradict each other: present both accounts side by side with their sources, marked "contested", instead of forcing one conclusion.
• Only one source found: mark the entry "unverified" and suggest queries that could cross-confirm it.
• Missing dates: file the entry under the earliest confirmable appearance and note the uncertainty.
• Regulated or expertise-bound domains: summarize public information only, avoid professional conclusions, and recommend consulting a qualified specialist.

Limitations and disclaimer

Conclusions depend entirely on public information reachable at runtime, so coverage is bounded by platform visibility, retrieval capability and language scope; omissions and lag are possible. Sentiment and stance figures are sampled summaries, not full-population statistics, and sampling bias is unavoidable. All trend reads are inferences from available information and do not constitute investment, legal, medical or other professional advice, nor any responsibility for decisions taken. Verify original sources before citing, and follow each platform's terms of use and information-handling rules.

Pro upgrade roadmap (Pay Skill reserved)

Billing will be enabled once the individual-creator payment channel launches. Free tier: single-topic single-pass research, basic fact table, condensed timeline, overall sentiment read, source list. Pro tier: parallel multi-topic research with side-by-side comparison, an industry-preset query matrix library, multi-dimensional sentiment crosstabs (audience x platform x time), continuous monthly tracking with period-over-period deltas, three report layouts (weekly / competitor brief / editorial planning), gap mining with an editorial calendar, batch export, and priority support. Upgrade path: switch to Pro in one click once the billing channel opens.

Standard 5-step workflow

1. Prepare input: gather requirement keywords and business context.
2. Configure parameters: set required parameters per this document (input, output format, constraints).
3. Execute: trigger the skill and follow the documented order.
4. Accept results: check output against the expected results and boundary conditions.
5. Reuse: save the successful template as a shortcut for next time.

Scenarios and triggers

Use this skill whenever you need, want, or are stuck on the following: High-frequency scenarios:

• Shipping a weekly industry report without trawling sources from scratch
• Sizing up what has happened in a target space before starting a new product
• Compiling an objective summary of a competitor's release and its user reception
• Separating overworked angles from genuinely uncovered ones in editorial planning
• Checking whether a spike in complaints is your product or an industry-wide pattern
• Collecting recent real cases and data points for a talk or internal share
• Catching up systematically on your field before an interview or performance review
• Reading what actual users are complaining about before a purchase or vendor selection
• Mapping how different parties interpret a newly issued rule and where they disagree
• Tracing the trigger event and spread path behind an unusual swing in brand mentions
• Judging whether a technical concept is still heating up or already cooling
• Writing a decision memo that needs sourced, re-checkable conclusions
• Running bilingual research for cross-border or multi-market work
• Reconstructing the propagation timeline of a public event
• Producing a monthly trend digest for a community or newsletter
• Deciding whether a topic deserves ongoing tracking resources Target users: product managers, market and brand researchers, content creators and editors, founders, analysts and consultants, and tech leads who must track their field continuously.