How to Find a Short-Story Topic: Match Ranking Samples to Your Strengths

Where Samples Come From, and the Limits of Each Source

$story-short-scan first asks which platform you want and whether you already have a direction. With a direction it scans that direction in depth; without one it surveys the whole list; for a comparison it compares platforms. The data source depends on the environment:

For Yanyan-style short fiction you will usually copy the samples yourself. Chart scanning here means finding patterns that repeat across a set of samples, not fixating on one title's rank.

No Collector? Copy the Ranking into a Sample File

When you supply a ranking, include title, author, genre or tags, word count, popularity (upvotes, saves or current readers all work) and blurb where you can. A bare list of titles can still be analysed, at one level lower confidence. The skill puts your material into the manual ranking format:

# 知乎盐言 · 热门榜

### #1 {title}

*{author} · {genre} · {status} · {count}字 · {number}收藏*

**标签：** {tag 1}、{tag 2}

> {blurb}

Keep the Chinese markers (字 = characters, 收藏 = saves, 标签 = tags), because the aggregation script reads them. Write popularity as number plus metric (the reference file's example is 2.3万收藏, 23,000 saves) so aggregation recognises it, and omit a missing field instead of inventing it.

Save the file in a new 扫榜/{YYYYMMDD}/ folder for this scan, then aggregate it into 扫榜聚合.md: for each platform and genre it lists count, share, median popularity and length, and frequent tags, and marks any genre with fewer than 10 titles as “少” (thin). Later analysis reads only this aggregate and pulls a few raw entries by genre or tag when it needs them.

What to Read in the Samples: Emotion Before Genre

The skill treats short fiction as an emotion market: readers finish one emotional experience in a short sitting, and a story spreads through completion and through shares, saves and likes. So besides genre names, the scan records frequent emotions, triggering situations, emotional peaks and the moments readers want to pass on.

From the aggregate it reads seven things: emotion types, recurring setups and situations, length, opening patterns, ending type (happy, sad or open), common title words and protagonist types, with openings and endings checked by sampling. On a Yanyan-style platform it can also look at the structure of highly upvoted stories, the genres and openings of less established authors who reach the list, which genres readers pay for, and how popular tags shift.

Three judgments run throughout. Emotional pull matters more than genre novelty. The first three sentences are the highest retention risk and must set up conflict, a status gap or an emotional hook. Reversal is a common sharing engine; without one, strong resonance, a strong talking point or a strong aftertaste has to make up for it. The conclusion is a report for the author that states sample size, sources and a suggested re-scan date, and leaves out script names and collection status.

Topic Matching: Strong Signals × Material You Can Sustain

After the conclusion is written, the skill rereads the conclusion file and the aggregate rather than relying on chat memory, then matches them against the project conditions you give. Priority goes to the intersection: a strong signal in the current samples that your material and abilities can support. Reversal and comeuppance stories have clear structures and are cheap to validate, so they count as low complexity; suspense and heartbreak need evidence of foreshadowing, reversal and emotional control, so they count as high complexity.

Each direction follows the topic-matching template with eight fields: target emotion, genre and setup, opening hook, sharing engine, complexity, feasibility, length and platform, and a low-cost validation step before drafting. Two or three directions are typical, most feasible first. Feasibility has hard caps: if the aggregate marks the group “少”, the sample has fewer than 10 titles, or it is only a list of titles you supplied, feasibility cannot be “high”; built-in-knowledge mode is always “medium”.

The result goes into 短篇扫榜结论.md, whose header records the scan date, the re-scan date (usually two to four weeks later) and the data source; a re-run replaces whole sections. /story-short-write reads this file when it sets the target emotion, so create your story project inside the scan folder or beside it, or else copy the file to the project root.

Original Example: A Former Housing Agent Scans a Ranking by Hand

The author, sample entries, titles and numbers below are invented for this article and do not represent any platform's actual ranking.

The author's conditions: five years as a second-hand housing agent, familiar with viewings, deposits and listing records; quick at family dialogue, weak at puzzles with several interlocking clue lines; wants to write a Yanyan-style short story of about 10,000 characters and has no direction yet.

The samples: from two screenshots of a popular list she copies 14 entries. All have genre and tags and most have a blurb; nine show a save count, five are cropped, so their popularity stays empty. One entry:

### #6 弟弟婚房写了谁的名

*青檐 · 家庭 · 已完结 · 11000字 · 4200收藏*

**标签：** 家庭、反转

> 爸妈说房子写弟弟的名字是为了好贷款，我在抽屉里翻到另一份协议。

(Title: “Whose Name Is on My Brother's Wedding Flat”; genre: family; tags: family, reversal. Blurb: my parents say the flat is in my brother's name to make the mortgage easier, then I find a different agreement in a drawer.)

After aggregation she groups the result by hand. The table is her note, not the script's output:

Every group has fewer than 10 titles and is marked “少”, so no direction can be rated “high”.

Topic matching fills in two directions using the template (illustrative). Direction 1, a property-sale reversal seen through an agent's eyes: the target emotion moves from anger to satisfaction with some warmth left over; the opening hook is her uncle asking her to underprice and rush-sell her grandmother's old flat while the system shows the grandmother herself listed it three months ago; the sharing engine is an old record reinterpreted; complexity low; feasibility medium (five samples in this direction, and her material can sustain it); validation is to write the blurb and the first three sentences. Direction 2, professional suspense about something wrong inside a flat during a viewing: complexity high, feasibility low, with only three samples and the weight falling on her weakness. The heartbreak group touches none of her material and is left out.

Her choice: Direction 2 is fresher, but she takes Direction 1. That is her decision; the matching only supplied an order and reasons.

First validation: she writes three sentences: “My uncle wanted Grandma's flat listed twenty percent under market, the sooner the better. I logged into the system: the flat had already been listed once, three months ago. The seller was Grandma, and the notes field held a single sentence.” Conflict, her double role and the hook all arrive within three sentences. She pulls the five raw entries in this direction and reads their blurbs; four start with a quarrel at a family dinner, so she keeps the listing-record entry point. Drafting her own blurb, she notices the satisfying ending relies only on Grandma scolding the uncle in public, so she changes it: acting as the agent, the narrator refuses the listing until Grandma confirms in person, and the uncle's plan stalls.

That is all that has happened. The story is unwritten and reader response is unknown; the re-scan is planned for three weeks later, when she will check whether this direction still shows up in the samples.

What the Result Decides, and Common Misuses

The result helps you rule out directions, rank candidates, record the basis and validation step for each one, and set a re-scan date. It does not choose for you and does not predict reads, saves, signing or income. For a platform it could not collect, the report gives the reason and says the conclusions exclude it.

Once you have a direction, use /story-short-analyze to dissect a same-direction story whose text you have, or start drafting with /story-short-write. If the story needs a longer arc, move to /story-long-scan. To turn the direction into a full piece, see turning an idea into a short story.

FAQ

Can I use the skill with no ranking at all?

Yes, but only through built-in knowledge. The skill loads its historical cross-platform reference, states in the report that the analysis rests on historical trend data and remains a candidate hypothesis until checked against a live ranking, lists the platform pages to re-scan, and rates every direction “medium”. Even a dozen or so copied entries with tags and popularity sit closer to the present, though confidence stays low.

How many samples are enough?

During aggregation, a genre with fewer than 10 titles is marked “少” (thin). If a direction is marked thin, the sample has fewer than 10 titles, or the samples are only a list of titles you supplied, feasibility cannot be “high”. A thin sample can still rule out directions and let you compare blurbs and openings, but it cannot be reported as a trend.

How soon should I scan again?

The re-scan date is usually two to four weeks after the scan; after that, the conclusion is reference only. A re-scan uses a new dated folder, compares against the previous aggregate, and replaces the “scan conclusion” and “topic matching” sections of the conclusion file in full.

Related reading

• 扫榜 — chart scanning→
• How to Turn an Idea into a Complete Short Story→
• How to Write Yanyan-Style Stories: Limited Viewpoint and Fair Twists→
• Choosing a Web-Fiction Genre: From Interest to a Trial Scene→
• Short Stories vs. Novels: Two Structures for One Idea→