aikey-title

This skill was originally named /zmm-…, renamed as of v1.0.0 to AI KEY（/aikey-…), and the old name will no longer be updated. Existing users upgrading, see README。

aikey-title: Title and cover formulas

First read aikey/references/家族公约.md(if unreadable → say plainly "the conventions cannot be read, so red lines cannot be guaranteed" and stop;do not substitute the conventions from memory), then read aikey/references/交互规范.md（🔴 reading it once does not count: when wrapping up, follow §4's three-part set — Recap · Before/After · next steps given as numbered options; when information is missing, follow §4 and usemultiple-choice questionsto ask,one question at a time; for cases where this does not apply, see §5), then read memory {vault}/08-技能记忆/aikey-title/ + _通用/, for theoretical sources see aikey/references/内容理论底座.md(when talking to the user, speak plain language only, do not mention theory names). The structure list and self-test method are in references/标题结构库.md。Note that it is not a viral formula library— empirical evidence shows the 12 title structures, across two accounts with different styles,not one reproduced across sources, so the so-called universal formula does not hold. (This refers to titles; for spoken-video openings there is another empirically strongest form, see aikey-hook。）

This skill's built-in criteria are in references/规则卡.md(criterion / why / how to check / strength); read it once before starting;{vault} when a corresponding rule file exists in the vault, the vault prevails and the rule card is the baseline.

Writing titles itself is worthless; any model can do it. This skill exists for only three reasons:

Choose the structure based on the shape of this content itself, not by applying a "most effective formula" (because none exists)
Clear the red lines— this is the layer generic tools lack
Explain why you chose this way, so the user can judge whether you are right

Every title must state which structure is used (S1–S12) and why it suits this content. If you cannot say it, it was written off the cuff.

Two ways of entering (distinguish first; output forms differ)

	Invoked directly by the user	Invoked as a routed call by the scriptwriting skill(added 2026-08-23)
What you have	A draft or a topic	The full body text, and it is not yet finalized
Output	Title candidates + structure numbers + red-line conclusion	Same as left,plus one more: whether the body text needs changes
Control	With you	With the caller, you hand it back once you have given the result

When invoked, do one more thing: push back

The cover headline and the first 2 seconds are two exits of the same decision. The cover says "stop adding features", and the first line says "stop adding features to your product"— these are not two separate things.

So when invoked, in addition to giving titles,One more thing to judge: given this title, does the opening of the body need to change?

Needs changing → say exactly which sentence to change and what to change it to, and hand it back to the caller
No change needed → also say so, so the caller knows you checked

🔴 This is the entire reason for introducing the call. Linear relay can't do this: by the time the relay happens the draft is already frozen, and the title can only accommodate the body.You were called in precisely so you can speak back the other way.

Hand off to the opening: three-line card (added 2026-10-10)

The title is set, next comes the opening (in the same call, or the user's next step goes to /aikey-hook），Hand over three lines, not a feeling：

① 方向：这条押的是哪一句（一句话，就是选中的那个标题想让人认出的东西）
② 边界：稿子 / 片子里实际有的材料，开头不能超出它（第 3–5 档时贴原句）
③ 兑现：标题的承诺在正文第几句还上

Three places say the same thing, but not the same sentence：

The cover big text giveskeywords, the title givesthe complete situation, the first spoken sentence givesa concrete scene or question
If any two of the three places have an identical whole sentence → change one of them. A repeated whole sentence means the audience reads the same sentence twice in the first 3 seconds, wasting the spot reserved for the gap

⚠️ When calledDo notgo through the full "first confirm which platform to publish on" Q&A — the caller already knows the platform. Produce directly, and save the Q&A for the user.

Where the title comes from (added 2026-08-23, precedes all formulas in this skill)

Before writing a title, first confirm where it comes from.

❌ Pick one off the viral hit list and rewrite it — that's copying a conclusion, and an identical title has no freshness ✅ Let itgrow outof the content: common sense → a phenomenon that can't be explained → deep theory → a counter-intuitive judgment → only after lifting it up one level does it become a title

Three criteria for a big cut; it counts only if all pass:

Even people who don't know this field know what you're talking about
It grows out of this piece of content's judgment, not picked from somewhere else
Taken on its own it still has a gap (they don't know what you're going to say, but they want to)

⚠️ Lift onlyone level. Lift two levels and you reach life philosophy, which is empty.

→ For the full five steps see aikey-topic/references/常识缺口法.md

🔴 Where the title comes from: from the landing sentence, not the viral hit list (added 2026-08-26)

First preserve the topic's unresolved question and the body's promise. If upstream follows aikey-topic/references/问题探寻与共鸣.md to get a concrete value, stay-or-go, investment, or feasibility question, you can polish it directly, with no requirement for emotional words to appear. The structure library is a classification of expression; it does not replace question generation, nor does it prove that a certain expression is effective.

One usable route: the emotional version of landing sentence ①, with the person changed(the following checks apply only to this route):

① 认知版   我的注意力每天都被外界安排        ← 认知词：注意力
                ↓ ★ 换情绪词（共鸣全在这一步）
① 情绪版   我每天焦虑什么，不是我自己决定的    ← 情绪词：焦虑
                ↓ 只换人称
标题       你每天焦虑什么，可能不是你自己决定的
                              ↑ 那是谁决定的？不说 —— 这就是缺口

🔴 If the emotional version is written right, the title is basically it with the person changed. If after changing it feels not punchy enough, the problem is in the emotional version, not the title.

🔴 The emotional version must contain a "price already paid" (added 2026-09-22, tested)

Same piece of content, same route, two outputs:

	The result of changing the person in the emotional version	Verdict
❌	"AI looks flashy, but what can it actually do for you"	Generic. It only describes confusion; after hearing it the audience feels no sense of loss
🟢	「Took a bunch of AI courses, but it seems nothing changed」	There's investment (learned a bunch), there's a gap (nothing changed), with sunk cost built in

🔑 The difference isn't in the sentence pattern, it's in whether there's something he has already handed over. Confusion is free,the cost is not — only someone who has already spent time or money will want to know why it was all for nothing.

🔴 Test (after writing the emotional version, ask one question)：

In this sentence, what has he already given up?

　Can answer (time / money / a pile of courses / half a year / sixty videos) = pass 　Cannot answer (only "don't know" "can't figure it out" "very lost") = still one layer short, go back and find that investment

⚠️ The cost must be something he has actually already paid, not a consequence you imagine for him(same as feedback_rules C13 / 裁定 #18）。

Verify two things before writing a title：

Semantic leap: is the emotional version still using cognitive words (think/judge/logic/information)? → go back and rewrite, don't force it on the title
Timeliness: is that emotion something heis enduring right now? ✅ anxious / uncertain inside / stuck / can't decide / can't sleep　❌ regret / detours / tuition (looking-back type)

⚠️ Field test (N20 v5): lifting from ① produced "The more you think, the more right you are, not because you thought right"—— "The more you think, the more right you are" is a cognitive phenomenon, not something anyone endures every day. No amount of title tweaking can save it; you have to go back and change ④.

🔴 If there's no landing line, don't write a title. Just pick a line from the content that sounds punchy, what you produce is a "sounds reasonable" title, not a "this is talking about me" title. 📄 aikey-topic/references/常识缺口法.md §④ Landing line

🔴 Read the main entry point first (added 2026-08-26)

{vault}/00-规则与索引/文案框架-总入口.md is the only navigation for "how to write one piece of copy": seven-layer map (stance→topic→structure→wording→red lines→process→medium) · conflict adjudication table · default drafting process · feedback auto-writeback mechanism。

🔴 Execute according to it by default; the user doesn't need to say "write according to xxx" every time. 🔴 Any feedback the user gives is automatically written back into the framework per its §5, without him having to point it out specifically. ⚠️ Follow its §6 checklistonly read the files needed for the current step, don't read all 20 at once.

Who it's spoken to

A solo knowledge creator who self-sources and self-edits. Three hard constraints (same origin as the commercial set):

Don't assume he has a team. Filming, editing, and publishing are all done by him alone; the plan must be executable by one person
No reporting upward. He makes the call himself; give judgments and actions directly, don't write "for reference"
Zero jargon. Use the theory, but don't let the terms appear——don't say "information gap", say "make him first feel he doesn't know"

Platforms are not one single standard

First distinguish the feed: single-column or double-column (added 2026-10-12)

Which feed the title is seen in determines what job it does.

	Single-column (Douyin, WeChat Channels)	Double-column (Xiaohongshu)
When the viewer sees the title	the video is already playing, he has already scrolled to it	You only see a thumbnail + one line of title,haven't clicked yet
The title's job	Help itstay, remember this one: the punchline, the core judgment, the viewing value	Compete for clicks with the image next to itClickbait
Can you state the conclusion	Yes. Giving the judgment doesn't count as a spoiler, but the body must have something beyond the judgment (why it holds, how it came about)	No. Once the answer is out, no one will click

⚠️ This one rewrites the scope of Gate 1 in the "5 Gates" below:"If reading the title makes you go 'So that's it,' send it back" only strictly holds for two-column layouts; on single-column platforms, giving the judgment directly in the title is a legitimate play.

Specs for each of the three platforms (default order: Douyin → WeChat Channels → Xiaohongshu)

	Douyin	WeChat Channels	Xiaohongshu
Post title	Recommended ≤20 characters, hard cap 30 characters (including punctuation), count it as soon as you draft. Prefer the punchline or judgment from the original text that can be passed along on its own	Recommended ≤20 characters (including punctuation). You can state the conclusion; preferphrasing suitable for forwarding on WeChat. When the content naturally points to family, colleagues, or clients,keep that sense of relationship	Capped at ≤20 characters, punctuation counts as characters(on the App, an emoji counts as 1 character, max 1). Specific scenario / contrast / number / question,don't give away the answer; it must carry its own context, so a stranger knows at a glance what it's about
Cover text	Two lines, each line ≤6 characters, total ≤12 characters (if the account has a tested template, follow the template).Top line: the subject or event; bottom line: the question or angle	Three lines, each line 3–6 characters, total 12–18 characters	Three lines, each line 3–6 characters, total 12–18 characters, each line a keyword phrase, not a long sentence
The cover's job	Homepage catalog index: accuracy matters more than attraction. From the cover alone, you should know exactly what this one is about	Compete for clicks: it also appears in chats, Moments, and forwarding cards. The cover shows what the audience wants to see; the title's job is to push it deeper or overturn it	Give a specific scenario, pain point, or result, leave the suspense to the title. If the event itself is strong enough, state it plainly; don't force in a twist
Post description	60–150 characters (up to 200 for complex ones); the first line goes straight into the conflict or scenario; keep only 2–3 key points; 1–3 topics (same rule card "Upload Three Fields": no industry jargon)	80–180 characters (max 220); in the middle, make clear "why it's worth forwarding to someone else"; at most 3 topics, none by default	180–350 characters; the first 2–3 lines give the conflict or a counterintuitive fact, the middle uses short paragraphs / numbering, keep one of the strongest cases; 3–5 topics, likewise no industry jargon

Cover text: the forbidden zone common to all three platforms:Do not use empty words like "truth / methodology / cognition / practical insights / underlying logic / subvert cognition" — words that could be used on any video mean nothing was said.

🔴 Before writing a Douyin cover, first answer this in your head: "This video specifically talks aboutwhom or what, and around it discusseswhat issue?" Compress the answer into two lines and that is the cover. If you cannot answer this, it means you have not yet thought clearly about what this video is about, so do not write the cover yet.

The user did not say which platform to post on → by default produce for all three platforms, in the order Douyin, Video Account, Xiaohongshu, and do not ask again. If the user only names one platform, produce only that one. For WeChat Official Account, Bilibili, Kuaishou, and other platforms,add them only if the user explicitly mentions them for that request。

⚠️ The three platforms musttruly be adapted: copying the same title / description three times with only the platform name changed = not done.

Red-line layer (this is the true core value of this skill)

Before writing, firstfilter out or rewritethese types of phrasing — many popular title styles are built on red-line landmines:

Sell truth: prohibitspecific income promisessuch as "earn over 10,000 a month / make 50,000 in 90 days / earn a million a year". You may keep the result-promise formula, but change it to a truth-abiding phrasing (see config.redlines.sell_real.allowed_phrasings). Public-domain titles must not contain specific income amounts.
Prohibited words: in titles, prohibit "most / first / only / whole internet" (prohibited-word avoidance table). In the structure library, whenever an example contains "most" or "only", replace it when using it ("the most fundamental reason" → "the real reason" "the overlooked reason"). This is not only about compliance, but also about effectiveness: two Douyin benchmark pools with different styles point in the same direction — pool A (n=541) titles containing absolute words have save_intent only 0.53×of the rest, pool B (n=259)0.72×(R2, empirical pattern library). Audiences themselves simply do not buy it. ⚠️ But this feature ispositively correlated on the person's own X account，Do not apply across platforms。
Do not name names: for phrasings that borrow a celebrity's influence, replace the celebrity with someone you can mention (you may mention OpenAI/Claude brands, but do not name people in config.redlines.no_naming.targets ).
Programming vocabulary ban(B-end oriented): covers / titles must not contain "programming / writing code / command line / programmer"; state the result, not the means ("let AI do the work for you" rather than "teach you to write code"). For exceptions, see Family Convention Six (read-aloud type / general-audience target topics).
Writing style: do not use English-colon-style titles (标题：副标 use Chinese full-width or rewrite), do not use English straight quotation marks ""(the standard is based on Family Convention §3.5; for quotation marks and emphasis use curly quotes “ ”).

Do not give formulas that fail the red lines. Before giving each one, run it through these five points yourself.

The 5 gates before a title goes out (all platforms)

After clicking in, is there still something to watch: if the title has already fully stated the conclusion, the body becomes a repeater. Self-test: after reading the title, is your mind saying "so that's how it is" or "and then what?" The former gets sent back.
- Single-platform (Douyin, Video Account) relaxed: The title can give the judgment directly; for self-testing, switch to "After reading the title, does he want to knowon what grounds?" If not = send it back. See "Split the information flow first" above.
Is the push pool big enough: Every insider term in the title (knowledge paid content / private domain / SaaS) filters out another batch of people who might be pushed to; switch to words that outsiders also feel (making money / gaining followers / not working a job / one person).
Does it poke a sore spot or a noun: "One-person company" is a noun, "being able to support yourself without a job" is the sore spot; "content automation" is a noun, "not wanting to rack your brains for topics every day" is the sore spot. What appears in the title should be the latter.
- 🔴 "Poking at other people's flaws" also doesn't count as poking a sore spot(added 2026-09-22, tested): "Those courses only teach you to type" pokes atthe course's flaw, while "I've taken a pile of courses but it seems like nothing has changed" pokes athis own situation。 The former requires him to first agree with your judgment about a third party before it's his turn to care about himself; the latter is him from the first second. Criterion:Is the subject of the title him, or the thing you want to criticize? If it's the latter = still one layer short, land it on him once more.
Is there pull: Contrast / specific numbers / suspense / conflict and controversy / time promise / result promise. Among these six kinds of pull, a title must have at least two; having only one is flat.
- "Specific numbers" holds only conditionally, it is not a universal item(R1, empirical pattern library): On a benchmark pool of n=541, titles with numbers were 1.73×, and numbers plus quantifier units were 2.33×; but on another independent pool with a different style (n=259)it did not replicate at all(0.88× / 1.09×) — that account simply doesn't use number titles. Correct reading: numbers + quantifiers arean accompanying feature of list-type / step-type / quantifiable-result-type content, not a universal rule.Use it only if this piece of content itself is that type; if not, don't force in numbers just to fit a formula(forcing them in may also hit the real red line about selling).
- "Result promise" ⚠️ not empirically verified: Pool A has no positive evidence (1.00×). You can use it, but don't treat it as a strength.
Check the character count: Xiaohongshu ≤20 characters including punctuation; if it exceeds that, cut it. (Platform display specification, not a tested conclusion.)
- Douyin cover: Based on the account's tested image-output template in {vault} ; when there is no template, use ≤6 characters per line. ⚠️ When a template exists, the template prevails — for example, one account's tested two-line template istop line ≤8 characters, bottom line ≤6 characters, and a one-size-fits-all 6-character rule would cut information that could fit.
- WeChat Channels / Xiaohongshu cover: Three lines, 3–6 characters per line, 12–18 characters total.

Ask three more questions: can this be passed on (added 2026-10-12)

The 5 gates govern "whether they click"; these three sentences govern "what happens after they finish reading":

Who will he forward it to? Can you name a specific person (wife, business partner, a friend currently signing up for a course) = there is a sharing action. Can't name one → see whether the original text has material that naturally points to a certain kind of relationship, and keep it in the title or description
In what future scenario will he think of this post again? A recurring life scenario (being pitched a course again, sitting through another weekly meeting) makes content more memorable than a one-off event
After it's passed around, do others still know who said it and what this person can offer? When the content has a story, is the author the role that can't be removed; if the title removes the author, then no matter how far it spreads, it's spreading on someone else's behalf

All three sentences can onlycome from material already in the original text. If the original text is just flat knowledge, go into the original text to find specific scenarios, numbers, contrasts, and consequences,don't invent a more sensational case。

Title, cover, and description each do their own job (added 2026-10-12)

	What it does	What it doesn't do
Title	Ask a question, create suspense, or nail down the viewpoint	Not responsible for summarizing the whole text
Cover text	Make clearthe specific object, scenario, conflict, or result	Not an abbreviation of the title
Description	The first sentence catches the strongest conflict; explain why it's worth continuing to watch; distill the key judgment or action checklist	Not pasting the verbatim transcript again; not changing it into an "AI summary voice"; not inventing data, experiences, identities, or results that aren't in the original text

When the cover and title appear at the same time, they shouldcomplement each other, saying the same piece of information twice wastes one slot. ⚠️ This is the default, not a prohibition: the account owner canactively chooseto use the same sentence for the title and cover (this has happened on the author's account).Choosing the same sentence after thinking it through is fine; forgetting to change it is not.

🔴 Write from the audience's side, not from the content's side (added 2026-09-22, this skill was truly defeated once)

For the same script, the skill gave three bets, the user casually offered one, and after evaluation the user's ranked first.

	Given by the skill	Given by the user
Title	"AI courses teach and teach, and all they teach you is typing"	《Why have I learned a bunch of AI courses but it seems like nothing has changed》
Where it grows from	The content's judgment(what's wrong with those courses)	The audience's situation(I invested, but nothing changed)
The first thing the audience needs to do	First agree with my judgment about a third party, only then does it become about them	Recognize themselves in the first second
Push pool	Medium (with a slightly judgmental tone)	Widest(zero insider words, everyone understands)

🔑 This isn't "he understands his own account better," it's a replicable positional gap:

The title stands on the content's side, and the audience has to take a step to reach it; standing on the audience's side, they don't have to move.

The final question before taking action (added after the 5 gates)

Is the subject of this title the audience's situation, or the thing I want to talk about?

It's the latter ⇒ you don't necessarily have to reject it, butyou must write another version that starts from the audience's end, then compare the two side by side. Don't only give the batch that grew out of the content end — that's where this skill has already been defeated once in real testing.

⚠️ An easy place to accidentally write it badly: whether to leave room

Two characters in the user's title are key:"seems"。

"Took a bunch of AI courses,nochange" = asserting he failed ⇒ he has to first admit he's no good before he can click in
"Took a bunch of AI courses,seems like nochange" = leaves room ⇒ he can click in without having to admit anything

🔴 Criterion：For this title to hold up, does he need to first claim an unflattering identity? Yes = add a softening verb (seems / appears / always feel like / is it that), or change the subject. 📌 This is the concrete way to apply two rules at the title level: don't make the audience first admit they're pitiful, and zero preaching throughout.

Workflow

Step 0: What stage is the video at (added 2026-10-10). Use references/拍摄进度分档.md to judge which tier this falls into: text only / not yet shot / shot and can pick up extra footage / shot and cannot pick up extra footage / cut from a long recording.If the context makes it clear, don't ask; if not, ask once with a multiple-choice question. The tier determines how much the title can promise:for a finished video, only promise what the video already says; for tiers 4 and 5, every candidate must mark which line in the corresponding video it comes from.

Step 1: Confirm topic + audience (ask for whichever is missing, one at a time); if the platform isn't specified, default to producing for all three: Douyin, WeChat Channels, and Xiaohongshu. Before starting, answer five things clearly in your head: thespecific objectbeing discussed, theaudiencemost easily hit, thestrongest hookin the original text (the most specific, most counterintuitive, or most emotionally striking fact / number / scene / line), theresultthe audience truly cares about, andwho he will forward it to. If the user pastes a piece of content → first extract the core topic and material (data / contrast / lines) before moving to Step 2.

Step 2: Choose the structure (start working). First judgewhat shape this piece of content itself is(list type? conventional-wisdom-overturning type? personal-experience type?), then pick 3–5 structures from S1–S12 thatmatch this shape, and write one candidate for each.

If there's already a title, it is candidate 0 (added 2026-10-10). Anything the user pasted, anything written in the script, anything posted last time, all count. Candidate 0 must also be labeled with a structure number and pass the red lines; if it has publishing data (clicks, completion rate, saves), include that too. After that, every new candidate must include one sentence onwhat point it is stronger than candidate 0 on(wider push pool / subject shifted to the audience's end / one more cost he has already paid...).If you can't state a specific point → bet on candidate 0, don't swap out a title that works just to look like you did something. Candidate 0 gets no exemption: when you have a landing line, it also has to be checked against the landing line to see whether it hits — if it doesn't, that's where the new candidate beats it.

Don't force a mismatched structure on just to pad the count— a piece about personal experience forced into "5 methods" means the title and content don't match, and people who click in will leave immediately.

Step 3: Generate + explain (stop). For each title: label the structure number (S1–S12) → say in one sentencewhy this structure fits this content → mark whether it passes the red lines and which spot was rewritten.

Don't cite any "original viral hit" as backing— someone else's viral hit can't prove your title will do well; that's exactly what this library overturns.

Step 4: 5 titles + 5 covers per platform, pick 1 top choice for each. The 5 candidates needgenuinely different angles(punchy line / essential judgment / concrete conflict / scenario / relationship / viewing value / action direction / outcome expectation…… cover at least 3 kinds per platform). Swapping synonyms or rearranging word order to pad the count doesn't count.The top choice must be picked by you, don't dump the whole picking job on the user; next to the top choice write one sentence on why you're betting on it, then note what each of the other two strongest backups trades for (some trade for a comment-section fight, some trade for clicks), so they can adjust by goal. Douyin's 5 covers aredifferent index phrasings of the same theme, pick the most accurate one as the top choice; WeChat Channels and Xiaohongshu's 5 covers should use different entry points.

Output format (first use no more than 5 lines to state the content judgment: core proposition, strongest hook, real pain point, who it'll be forwarded to, in what scenario it'll be remembered later; then output in platform order):

## 话题：[x]  ｜ 人群：[x]  ｜ 档位：[第几档]

### 一、抖音
| # | 标题 | 结构 | 角度 | 为什么配这条 | 红线 | 推荐 |
|---|---|---|---|---|---|---|
| 0 | [已有标题，没有就删这行] | S[n] | … | … | ✓ / 改了哪 | |
| 1–5 | … | S[n] | 金句 / 判断 / 冲突 … | 一句话 | ✓ | 首选 / 备选 |
**首选标题**：[…] —— 为什么押它；备选 [n] 换来 …
**作品描述**：[一版，可直接复制] ＋ 话题 1–3 个
| # | 上行（对象 / 事件） | 下行（问题 / 角度） | 推荐 |
**首选封面**：
[上行]
[下行]
**分工**：1–2 句说标题、描述、封面各管什么

### 二、视频号
（同上；封面三行；分工那句改说为什么适合在微信里转）

### 三、小红书
（同上；封面三行；分工那句改说标题和封面怎么一起抢点击）

Special cases

Optimizing an existing title: treat the original title as Candidate 0 per Step 2 → first judge which structure it uses (even if the user doesn't know) → judge whether that structure fits this content → switch structures, or revise the specific wording within the same structure.If the newly written one can't say where it beats it, the conclusion is "the original title doesn't need changing", and that's a valid conclusion.
Just want to browse what phrasings exist: give the S1–S12 quick-reference table (see the structure library),and make clear it's a vocabulary list, not a ranking, no one is inherently better.
Topic not decided yet: work backward from the audience instead, ask what first reaction they want the people who see it to have (want to know why / feel a jolt / "this is so me" / want to learn), and use that reaction to work backward to which kind of structure to use.

Absolutely never do

Don't give titles that can't state a structure number and reason.
Don't claim a certain structure is "more likely to go viral"— the 12 structures have zero cross-source replication, so saying that is making it up. To give an effect judgment, it can only be based on the user's own data (scripts/title_struct.py can run).
Don't give titles that fail the real-selling / banned words / no naming / programming vocabulary red lines.
Don't change the underlying logic of the formula (the core of "why X is actually harmful" is counterintuitive, it can't be changed to something not counterintuitive).
Don't cram it in when it's over the word count.

Memory

Self-check before finishing: which type of title style did the user reject (record under "Corrections", e.g. "Don't use fear-based anxiety-inducing ones")? Which formula / cover method was validated as effective by /aikey-retro's click-through / completion-rate data (record under "Effective methods", with values + platform)? Write to 08-技能记忆/aikey-title/, and check for duplicates first.

Don't know the next step → go back to /aikey。