What to Collect: Source Priority and Fields to Record
First settle two things: which platform, and whether you already have a direction in mind. With a direction, dig into it; without one, start with an overview of the whole chart; to compare platforms, scan each one separately.
| Priority | Sample source | Use and limits |
|---|---|---|
| 1 | Platform collector scripts | Need network access. Qidian works without Chrome by default; Fanqie, Jinjiang, Qimao and Ciweimao need Chrome launched with /browser-cdp first. A redesigned page or a verification wall can stop collection |
| 2 | Rankings the author supplies | Paste text, screenshots or links; they are put into one format and aggregated the same way |
| 3 | Built-in knowledge | Only when there is no network and no data. The report says this is historical trend data, usable only as a candidate hypothesis until checked against live charts, and lists the charts to rescan |
New-book charts give early signals of new genres and shifts; reading, monthly-ticket, best-seller and “gold” charts show traffic and paid support that already exist. Take both kinds. For each title, record at least the chart name, date, position, title, genre, status, word count, heat with its unit (readers, favorites or monthly tickets) and the tags at the start of the blurb. Save each scan as its own dated record rather than appending to an old one. If a chart could not be collected, say the conclusion leaves it out.
A Single Rank Is a Lead; Repetition Is the Signal
A chart proves only that a book is there, not how strong demand is. The rule in story-long-scan: one title on a chart is an isolated case; only when similar samples repeat across charts and authors does a pattern become a trend candidate.
- Enough samples? If a direction has fewer than 15 titles (fewer than 10 on a small platform such as Ciweimao), the aggregate marks it “few”, meaning sparse data, and feasibility cannot later be rated high.
- Spread wide enough? A combination seen on one chart or from one author is a lead. Appearing on both new-book and reading-type charts is more persuasive than one chart.
- Same unit? Fanqie measures traffic and completion, Qidian subscriptions and ongoing readership, Jinjiang favorites and points. Do not compare numbers across platforms or across charts.
Then grade anything new. New genres are rare and count as a trend only with repeated samples across charts and authors. New golden fingers (protagonist advantages) are mostly refreshes; they suit low-cost tests but need a different genre shell and emotional trigger. New gimmicks are the most common and carry a high risk of sameness; without a differentiating mechanism, lower their priority.
Sort Every Conclusion: Observation, Lead, Hypothesis, Next Check
In the aggregate, look in turn at genre distribution and counts, word-count ranges, common title words, hot tags and the opening hooks in representative blurbs. Then pull a few raw entries for a candidate direction to spot new character setups, opening angles and set pieces. Write “rising”, “steady” or “falling” only if you have an earlier scan to compare; otherwise describe the present. Then put each conclusion in one column:
| Column | What goes in | Rule |
|---|---|---|
| Observation | A repeated pattern that several samples support | Say how many titles and which charts |
| Lead | A new combination seen in one title or on one chart | Note it; check again next scan |
| Hypothesis | Why readers follow it; whether your version can work | Mark it “to verify”, never “will be a hit” |
| Next check | A concrete action that confirms or refutes it | More samples, a breakdown of a representative book, a trial opening |
When studying blurbs in the same lane, take only the functional slots: the problem, the benefit, a preview of the payoff, reassurance and suspense. Do not copy specific plot or wording.
Reader Profile: Write What the Samples Support, Mark the Rest Unknown
The profile uses nine dimensions: platform, tags, reading volume, gender, age, region, education, class and social role. Tags set expectations: a “system” tag promises a levelling rhythm. Class decides which kind of comeback resonates; social role decides which daily life feels like the reader’s own.
For resonant material, use 70/20/10: 70% experiences readers have had, such as generational memories, school and childhood; 20% pressures they face now, such as work, relationships and money; 10% current hot topics. For every dimension, note which samples support it. Where the samples say nothing, write “unknown” instead of filling in an impression.
Original Example: From a Hand-Copied Sample Table to a Direction Worth Testing
Everything here is invented. Codes such as A1 replace titles and match no real book, position or platform data. The table is an imaginary author’s hand-made sheet, not the skill’s output.
The author spent a few years doing house calls for the family’s appliance-repair shop in a county town and wants to write a long novel of everyday urban life for a platform already chosen. With no collector set up, the author copies 8 entries in this direction by hand from the new-book and reading charts, noting chart names and dates in the header. The table keeps only the columns this judgment needs.
| No. | Chart | Blurb tags | Opening angle | Reward pattern |
|---|---|---|---|---|
| A1 | New-book | business, daily life | Jobless and back home, takes over a failing car-repair shop | Each fixed job unlocks a regular customer |
| A2 | New-book | business, system | Takes over grandmother’s breakfast stall, behind on rent on opening day | Good reviews earn rewards |
| A3 | New-book | daily life, craft | Laid off, takes over a watch-repair shop on an old street | Each repaired watch brings a referral |
| A4 | Reading | farming, food | Goes back to the village to lease a fish pond | Daily check-in rewards |
| A5 | Reading | business, system | Takes over a key-cutting stall left by a mentor | Each opened lock earns a reward |
| A6 | Reading | urban, system | Leaves the army and opens a farm guesthouse | Quest rewards |
| A7 | New-book | business, mystery | Takes over an uncle’s junk shop; the first deal brings a box of unknown origin | Old objects lead to old cases |
| A8 | Reading | daily life, ensemble | Takes over a hardware shop in an urban village | No golden finger, only neighbors’ goodwill |
| Column | Content | Basis |
|---|---|---|
| Observation | Most openings take over an existing shop or stall, on both charts | A1, A2, A3, A5, A7, A8 |
| Observation | Rewards come fast: finishing a job or task brings a reward or a regular | A1, A2, A3, A5, A6 |
| Lead | Opening pressure from job loss, layoffs or overdue rent | New-book chart only: A1–A3 |
| Lead | Business plus mystery; goodwill instead of a golden finger | One title each: A7, A8 |
| Hypothesis | Readers follow for the reliable “one job, one payoff” return and a growing circle of familiar faces | To verify by breaking down books |
The profile covers only what the samples support. Tags lean to business and daily life, so readers expect each job to resolve. Hometowns, old streets and urban villages can feed the 70% of nostalgic material. The 20% of current pressure rests on three new-book entries only, so it stays a hypothesis. Hot topics, gender, age and education have no basis and stay blank or “unknown”.
Candidate direction: a county-town repair shop where every house call to fix an old appliance pulls the repairer into one household’s trouble.
- Genre mix: urban daily life plus business. Following “85% familiar, 15% surprise”, keep the familiar takeover opening and put the novelty in the house calls
- Reward: each repaired job brings a returning customer or a lead to the next house call
- Difference: in the samples the shop is the stage; here the conflict happens in the customer’s home, and the parts, old models and house-call etiquette come from the author’s own work
- Feasibility: medium. Only 8 samples in this direction, so the aggregate would mark it “few”; with no earlier scan, the trend can only describe the present; the material covers the opening, but whether it can carry a long book is untested
- Failure risks: takeover openings are crowded, so a first chapter that only covers the takeover will sink; without a system, rewards may feel slow, and A8 is the only sample that clearly has no golden finger; house-call units can repeat as “fixed it, got thanked”
- Validation (planned, not yet done): on the next scan, bring this direction to 15 or more samples and look for system-free titles and counterexamples that flopped or were abandoned; break down the first three chapters of A5 and A8 (one with a system, one without) to test the “one job, one payoff” hypothesis; list house-call units and check whether later ones still change the conflict rather than just the appliance
This card is only a candidate: the author may add a light system, lean toward mystery or drop it. Until the sample is large enough, the skill’s rules do not allow a high rating.
How the Scan Feeds a Topic Decision Without Making It for You
The scan report is written for the author, covering market conclusions and directions you could write, and is saved to 选题决策.md (the topic-decision file) in this scan’s folder. A scan that stops short of choosing a topic still leaves its findings on disk.
When you want topics, the skill first asks for your target platform, existing material, strengths or writing constraints, and planned length. Each recommended topic then goes through four steps:
- Why it could take off, as a hypothesis: inferred only from repeated samples and marked “to verify by breakdown”.
- Market check: how many titles share the direction, the trend, and any counterexamples that flopped or were abandoned.
- Differentiation: match your strengths to a gap in the market and say how your version differs from what is on the chart.
- Feasibility, failure risk and validation: no “high” without enough samples, and “medium” for every direction when only built-in knowledge was used. A common check is to write the first three chapters, see whether readers keep following, and switch if they do not.
It recommends two or three and leaves out directions your material cannot support. After a book breakdown, the “why it could take off” line gets filled in. When you start the book, story-long-write can read this file as its starting point; if the project sits in the same folder or next to it, it finds the file and asks you to confirm. Which one to write, how to change it and whether to drop them all is your decision.
FAQ
Can I scan charts without a collector or Chrome?
Yes. Give the skill the chart text, screenshots or links and it puts them into one format and aggregates them; you can also copy entries by hand, as in the example. Only with neither does it fall back to built-in knowledge, which the report labels a candidate hypothesis, with every direction rated medium.
How many titles are enough?
The skill’s threshold is 15 titles per direction, or 10 on a small platform such as Ciweimao. Below that the aggregate marks it “few” and feasibility tops out at medium. Reaching the number only means the sample is not sparse; the trend, counterexamples and your own material still decide.
How is chart scanning different from breaking down a book?
Scanning looks across many titles for repeated patterns and candidate directions; a breakdown reads one book closely for textual evidence and transferable mechanics. The “why it could take off” that a scan produces is a hypothesis, filled in by breaking down representative books; see how to break down a novel.
Do it with the skill
Use $story-long-scan to scan this direction on the platform I chose. Use an available collector first, otherwise the rankings I supply, and built-in knowledge only as a labeled fallback. Treat a single title as a lead; split conclusions into observations, hypotheses and next checks, with sample counts and source charts. With fewer than 15 titles in a direction (10 on a small platform), cap feasibility at medium. Then suggest two or three candidate directions that fit my material, planned length and target platform, each with failure risks and a validation step. I choose; do not write prose.
More of the method: Nine-dimension reader profile and 70/20/10 · Grading new genres, golden fingers and gimmicks