social_trend_research v1.0.0

---

name: social_trend_research description: Find social trends worth acting on: track rising topics, sounds, and formats in a niche, judge where each trend sits in its life cycle, run brand fit and rights checks, and turn the survivors into dated content briefs. Use when someone asks what is trending in their niche, whether a trend is still worth joining, to build a weekly trend sweep, or to turn a found trend into a post before its window closes.

---

Social Trend Research

Purpose

Most trend work fails in one of two places. People chase a list of popular posts with no check on timing, so they publish at the peak. Or they find a real rising trend and lose the window while the brief sits unfinished.

This skill runs the full path: find candidates from platform native sources, confirm each one on a second independent source, place it in its life cycle, score it, run hard fit and rights checks, and hand off a brief with a publish deadline. It is source neutral. No paid scraping tool is required, and every step can be rerun from free platform surfaces listed in references/source-map.md.

Workflow

1. Set the sweep brief

Write down before you search:

• Niche and audience, in the words the audience uses for itself.
• Platforms where you will actually publish, one to three, not all of them.
• Content types you can make this week, with the gear and people you have.
• Topics, claims, and joke styles that are off limits for this account.
• Sweep window: daily for fast sound and format trends, weekly for topics, monthly for seasonal planning.

2. Collect candidates from native sources

Work the sources in references/source-map.md. Start with the trend surface of the platform you will publish on, for example TikTok Creative Center for sounds and hashtags, YouTube search and trending for formats, Google Trends for topic demand, Pinterest Trends for seasonal and lifestyle planning, and niche subreddits and mid size creators for early language.

Log every candidate in assets/candidate-log.csv format: name, type (topic, sound, format, phrase, or aesthetic), first source, date seen, and three example posts by different creators.

Collect 8 to 15 candidates, then stop collecting. A longer list does not improve the two or three posts you can actually ship this week.

3. Confirm on a second source

Require a second, independent confirmation before scoring:

• A TikTok sound rising in Creative Center plus recent uses by niche creators you follow, or a matching rise in Google Trends related queries.
• A phrase repeating in comments across several creators plus platform autocomplete now suggesting it.
• A topic rising in Google Trends, where trending status reflects both a relative spike and absolute search volume, plus posts appearing in the niche community itself.

Single source candidates stay in the log as WATCH items. Bots, one paid push, or one large account can fake a single feed. They rarely fake two different systems at once.

Read recency, not totals. Google Trends data is normalized to 0 to 100, with 100 as the peak for that query set, so it compares shape and timing, never absolute volume. On a sound page, recent uses in the last 48 hours matter more than a total use count built over months.

4. Place the trend in its life cycle

Use the five stages in references/lifecycle-and-fit.md: signal, rising, peak or saturating, declining, and evergreen conversion. The stages adapt Rogers adopter groups and the Gartner hype cycle shape to social speed, where a format can move from signal to peak in under a week.

Write the stage and the evidence for it in the log. Evidence means dated observations, such as newest uses are two hours old and led by mid size creators, not a feeling that a trend seems big.

5. Score the candidate

Fill a JSON file for each serious candidate and run the scorer:

bash
python3 scripts/score_trend.py candidate.json
python3 scripts/score_trend.py --demo

Score each field 0 to 5 honestly, from the evidence you logged:

• velocity: growth over your window against the prior window.
• acceleration: growth speeding up, holding, or slowing.
• breadth: distinct creators and communities using it.
• cross_platform: independent sources confirming it, 0 means one feed only.
• niche_relevance: overlap with the audience you serve.
• production_fit: can you make a native version in your normal format, fast.
• longevity: can the structure survive as a series or evergreen format.

Add days_since_first_seen, and set rights_blocked, sensitive_origin, or brand_conflict to true when they apply.

The scorer returns momentum, fit, and total scores out of 35, a stage, and a verdict: ACT_NOW, TEST_SMALL, WATCH, ADAPT_EVERGREEN, REWORK_OR_SKIP, or SKIP. Treat the verdict as triage. A human still reads the warnings, in particular the single source warning and the late window warning at peak stage.

6. Run the hard fit checks

Run the checks in references/lifecycle-and-fit.md in order. Any hard failure blocks the brief, whatever the score says:

• Rights: for commercial posts on TikTok, the sound must come from the Commercial Music Library, which TikTok describes as a library of 1 million pre cleared songs for business use on TikTok, or you must hold a license that covers commercial use on that platform. A track cleared for one platform is not cleared for another.
• Origin: search the format name plus origin. Formats built on a tragedy, a death, or a joke at a group's expense stay blocked even after they look harmless.
• Claims: health, money, and result claims inside a trend need proof for your version, or the claim comes out.
• Brand conflict: a joke that mocks your own customers, reverses a position the account took on the record, or fakes product experience is a block.

Soft checks decide the twist: state in one sentence what your version adds that the top three examples do not already do. No twist at peak stage means skip it.

7. Write the brief and set the deadline

Use assets/trend-brief-template.md for every ACT_NOW and TEST_SMALL verdict. The brief carries the evidence, the rights checker name, the hook, the beat by beat structure, the caption draft, any disclosure needed, the owner, and the publish deadline.

Deadlines by stage: rising, publish inside 48 hours. Peak with a twist, publish inside 24 hours or pass. Signal tests, publish cheap and fast, and log the result either way.

If the post involves a material connection to a brand, payment, free product, employment, or a financial tie, the brief must include a clear disclosure in plain words such as ad or sponsored, placed where viewers see it without expanding the caption. That follows FTC endorsement guidance for social posts. Vague tags such as #sp or #collab do not meet that bar.

8. Measure and convert

Check each posted trend piece at 48 hours: saves, shares, comments with substance, and profile actions. Views alone do not tell you if the trend fit the audience.

Decide from the pre written rule in the brief: repeat, adapt, or drop. When a format worked but the reference has dated, file the structure under evergreen conversion and pitch it as a series entry. That conversion step is where trend work pays off after the window closes, most teams skip it and start from zero each week.

Close the loop in the candidate log: final verdict, post link, 48 hour result, and what you would do earlier or later next time.

Worked example

Candidate: a rising sound used for short tutorial reveals in a home cooking niche.

Evidence: Creative Center lists the sound as rising for the region over 7 days. Three niche creators with mid size followings posted uses in the last two days. Google Trends related queries for the recipe topic show the phrase from the sound as a rising query. Newest uses are hours old.

Score: velocity 4, acceleration 4, breadth 3, cross_platform 3, niche_relevance 5, production_fit 4, longevity 3, days_since_first_seen 3, no blockers. The scorer demo uses these exact numbers and returns total 26 out of 35, stage rising, verdict ACT_NOW. Run it with python3 scripts/score_trend.py --demo to see the full output.

Rights twist that saves the post: the chart original is not cleared for commercial use, so the brief specifies the Commercial Music Library version or a licensed cover with the same structure, and notes the checker name. The brief sets a 36 hour deadline, hook is the finished dish in the first two seconds, then the reveal steps timed to the sound. The 48 hour rule set before posting: repeat as a series if saves beat the account median, drop it if comments call it dated.

Counter example: the same sound three weeks later, 2 million total uses, brand accounts leading, comments repeating the joke back. Velocity and acceleration fall, stage reads peak or saturating, and the honest verdict is skip it as a trend play and log the reveal structure for an evergreen series instead.

Output Contract

Every sweep returns:

1. The candidate log, with source, date, stage evidence, score, and verdict for each candidate, including the ones you skipped and why.
2. Scorer output for each scored candidate, with blockers and warnings kept in the record.
3. A filled brief in the template format for each ACT_NOW and TEST_SMALL candidate, with a named owner and a publish deadline.
4. A 48 hour follow up entry for each posted piece, with the repeat, adapt, or drop decision.

Operating Rules

• Never call a trend rising from one source, one post, or one vendor dashboard. Two independent sources, dated, or it stays a watch item.
• Never quote a platform number without its window and region. A use count, a Trends index value, and a survey result all mean different things, and each source file note in the scratch research states its limits.
• Never invent a trend origin, a first use date, or a creator name. If origin cannot be verified, write origin unverified and treat the origin check as failed for commercial work.
• Never brief a blocked candidate to keep a schedule filled. A skipped week beats a rights strike, a muted post, or a joke with a bad origin.
• Keep the twist honest. Do not fake experience with a product, a place, or a result to fit a format. FTC guidance treats endorsements as statements the audience trusts, disclose material ties and only claim what you did.
• Respect platform terms and access limits. Trend surfaces show public posts. Do not scrape private groups, bypass logins, or promise coverage of accounts you cannot see.
• Update the source map when a platform moves or renames a trend surface. Trend tools change often, a stale link in a brief wastes the window the brief was meant to save.