模型未返回最终正文（finish_reason=length），以下为推理过程摘录，可尝试提高 max_tokens 或改用多段生成。

我们需要写一个HTML片段，直接写入数据库。要求严格。需要输出正文片段，无 markdown 围栏，无解释。允许的标签只有 p 和 h3。禁止 img，script/style。结构顺序：第一个元素 p：依据首段模板改写应用介绍，必须自然出现“冒险达人”。然后按小节顺序输出：软件介绍、操作指南、推荐理由、使用说明、特色、测评。但注意整体说明：“然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）”

然后下面【小节顺序】是：软件介绍 → 操作指南 → 推荐理由 → 使用说明 → 特色 → 测评

但是“随机抽 3～6 个非测评小节” + 末位测评 = 4～7 h3。然而后面又给每个小节任务，包括全部六个。似乎需要选择一部分？但用户列出的小节顺序全部，且每个小节都有具体条数要求。仔细看 “然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>（标题须与下列 {H3title} 完全一致），非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**（本次要求：用序号分条：每条单独成行，行首为「数字+英文句点+空格」（示例 1. 2. 3.），全篇保持同一格式）；**每节要点条数以该小节下方【条数】为准**（在 3～6 条之间且各节独立）。**

需要理解：用户给定所有小节顺序，但开头说“已从词池随机抽 3～6 个非测评小节”。这意味着我们不应该输出全部非测评小节？但是后面“小节顺序”却完整列出六项，并且每个小节都有任务说明。可能存在冲突。需要进一步解读。

用户说：
【整体结构】顺序必须严格如下：
1) 第一个元素是一个 <p>：依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍。
2) 然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>（标题须与下列 {H3title} 完全一致），非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**（本次要求：用序号分条：每条单独成行，行首为「数字+英文句点+空格」（示例 1. 2. 3.），全篇保持同一格式）；**每节要点条数以该小节下方【条数】为准**（在 3～6 条之间且各节独立）。
3) **最后一个小节固定为「冒险达人测评」**（已在下列顺序末位）：**仅 1 个** <p>，**不要**序号与分条，**一段**短评，**全文不超过 200 字**（含标点）。

然后是【小节顺序】软件介绍 → 操作指南 → 推荐理由 → 使用说明 → 特色 → 测评

然后是每个小节任务，每个都給条数。看起来像是希望全部输出这些小节？但开头说随机抽 3～6 个非测评小节。如何调和？

可能意思是：在生成时，从词池中随机抽3-6个非测评小节，但这里实际已经给出了“小节顺序”且每个小节都有任务，也许所有的小节都是随机抽出的？但“随机抽3～6”与“全部六个”矛盾。也许“已从词池随机抽 3～6 个非测评小节”是在描述系统的过程，但最终输出需要“仅按下列【小节顺序】”即所有列出的? Wait "仅按下列「小节顺序」输出" 后面有完整顺序，所以似乎应该输出全部列出的小节？ 然而全文 h3 数量是 4-7，如果输出全部六个非测评加一个测评，共 7 个 h3，符合 4-7 的范围内（7）。所以可以输出全部。同时开头“已从词池随机抽 3～6 个非测评小节”也许不影响，因为已经选择了6个（包含全部）? 6 在 3-6 之间。全文 h3=7，符合。所以安全输出全部六个非测评+测评。每个 h3 后按照给定条数写。总共 p 数量：首段 p=1 + 软件介绍3 + 操作指南3 + 推荐理由5 + 使用说明3 + 特色6 + 测评1 = 22 个 p。这个不算离奇。

但需要注意：输出必须是可直接写入数据库的 HTML 正文片段。禁止 markdown 代码围栏。所以不要用 ```。

需要 ensure all text in Chinese. 需要自然出现应用名称“冒险达人” in first paragraph. Also each h3 title exactly as specified: <h3>冒险达人软件介绍</h3>, <h3>冒险达人操作指南</h3>, <h3>冒险达人推荐理由</h3>, <h3>冒险达人使用说明</h3>, <h3>冒险达人特色</h3>, <h3>冒险达人测评</h3>. Good.

Need serial numbers: 全篇统一为同一种序号格式：每条单独成行，行首为“数字+英文句点+空格” (e.g., 1. 2. 3.). For each section independent counts as specified. Since each <p> contains one numbered point, each point begins with number. Need numbers reset per section? Format "全篇统一为同一种序号格式" not necessarily continuous across sections. In examples, each section begins with 1. Usually each small section has 条数. The instruction: "每节要点条数...各节独立" implies reset per section; each section's first point is 1. So do that.

Need ensure each point sentence within 50 chars. Let's craft carefully.

First paragraph: Must be original rewrite of first段, include app name, combine logic. App Store intro style. Should mention: 冒险达人是一款集多种娱乐项目于一体的应用，提供娃娃机、推币机等玩法，无需去商场，手机实时操控娃娃机，正侧面双摄像头高清直播无死角，精准抓取，超强抓力提升成功率. Need “冒险达人” in first p. Use maybe: 冒险达人是一款集合娃娃机、推币机等丰富玩法的线上娱乐应用，让用户无需前往商场，即可通过手机实时操控真实设备；借助正侧面双摄像头高清直播与无死角视角，能精准抓取娃娃，超强抓力大幅提升抓取成功率。 That is one p. Good.

Software Introduction section, 3 points, each ~50 chars. Need product positioning, target users, overall capabilities. Let's craft:

<h3>冒险达人软件介绍</h3>
<p>1. 冒险达人面向喜爱抓娃娃与休闲游艺的用户，提供一站式线上娱乐体验。</p>
<p>2. 内置娃娃机、推币机等丰富玩法，覆盖多种线下经典游艺项目。</p>
<p>3. 用户可远程实时操控设备，用高清直播与多角度视野完成精准抓取。</p>

Check each <50 chars? Let's count: 1. 冒险达人面向喜爱抓娃娃与休闲游艺的用户，提供一站式线上娱乐体验。 (roughly 31 Chinese chars) fine. No issue.

Operation Guide 3 points: key flow steps. Need no "1." continuity with previous? It's okay reset. Each <p> one point. Could be:

<h3>冒险达人操作指南</h3>
<p>1. 打开应用后选择娃娃机或推币机，进入在线设备房间即可启动直播画面。</p>
<p>2. 通过屏幕摇杆控制机器抓手，结合正侧面双摄像头调整角度并确认目标。</p>
<p>3. 点击抓取按钮实时操作，若成功抓中，可在线查看战利品与发货信息。</p>

Need each ~50. Fine.

Recommendation reasons 5 points: why install, benefits. E.g.

<h3>冒险达人推荐理由</h3>
<p>1. 无需前往商场，在家即可随时畅玩娃娃机和推币机等热门游艺。</p>
<p>2. 双摄像头高清直播提供正侧面无死角视野，抓取位置看得更清楚。</p>
<p>3. 超强抓力提升抓取成功率，减少反复尝试带来的挫败感。</p>
<p>4. 多种玩法集中在一个应用，满足休闲娱乐和收集奖励的多元需求。</p>
<p>5. 实时操控体验真实有趣，适合碎片时间放松和好友分享比拼。</p>

Need each point 50 chars. Fine.

Usage Instructions 3 points: installation, permissions, account, basics. Need original based on first segment (not too much info). Craft:

<h3>冒险达人使用说明</h3>
<p>1. 下载安装后在首次启动时同意摄像头与网络权限，确保直播操控流畅。</p>
<p>2. 注册并登录账号，充值或领取体验币后即可选择设备开始游玩。</p>
<p>3. 建议在稳定网络环境下使用，并根据画面提示调整手机方向和操作手势。</p>

Good.

Feature section 6 points: relative diff. Need "相对同类软件最突出的差异能力与使用场景". Based on original. Need six points, each 50ish. We can cover: real device remote control, dual camera live, no死角, strong grip, multiple play, no need mall, etc. Need avoid duplicates from previous. But can rephrase emphasizing features. Let's craft:

<h3>冒险达人特色</h3>
<p>1. 真实设备在线操控，手指轻点即可远程玩线下同款娃娃机。</p>
<p>2. 正侧面双摄像头同步直播，抓取视角更全面，减少盲区误判。</p>
<p>3. 高清画面配合实时回传，无需到店也能感受现场游玩氛围。</p>
<p>4. 超强抓力设计有效提升出爪成功率，让奖励更容易到手。</p>
<p>5. 娃娃机、推币机等多项玩法一站聚合，切换选择更轻松。</p>
<p>6. 随时打开应用即可开玩，通勤、居家等碎片场景都适合体验。</p>

Check all are separate and within 50. Fine.

Survey section: only one <p>, no numbers, continuous short review, total not exceeding 200 characters. Need objective and concise, personified maybe. Need no numbering, avoid numbers. Should "冒险达人测评" as h3 then p. Need total paragraph text <=200 Chinese chars including punctuation. Let's craft a concise assessment ~130-180 chars. Sample:

“冒险达人把线下游艺厅常见的娃娃机、推币机搬到手机上，用双摄像头直播和远程操控解决了“看不见、抓不准”的痛点。对喜欢抓娃娃又没时间逛商场的人来说，它提供了一种低门槛的新选择。超强抓力提升了成就感，多种玩法也增加了留存乐趣；若后续在设备排队和发货体验上持续优化，会更值得一试。”

Need count maybe around 140. We'll ensure no "1." etc. Good.

Need ensure full output is just HTML fragment, no markdown fence. Need maybe no newline before/after? Fine.

But let's think more deeply: Are there any hidden constraints about "应用名称（须在首段 <p> 正文中自然出现）" Yes we put in first p. First p must be first element. Good.

Need no <img>, no script/style. Good.

Potential issue: The user said "全文共 4～7 个 <h3>" and "已从词池随机抽 3～6 个非测评小节". If we output all 6 h3, total h3=7, within 4-7. So okay. But maybe "仅按下列小节顺序" means only those listed after "已随机抽" that are chosen, not all possible. Yet they listed all possible six, which is 6 non-测评 -> h3 count 7, valid. No conflict. Good.

Potential issue: "然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）" Parenthetical states "已从词池随机抽3~6个非测评小节" but "下列" matches the overall order. It says 若含“介绍”类关键词会固定排在第一. Since we include 软件介绍 as first h3. OK.

Need all H3 titles exactly. We have them.

Need each section uses same numbering style. Yes.

Need each "点" alone on line? In HTML, `<p>` each point already separate. Good.

Need ensure "每节要点条数以该小节下方【条数】为准" We have counts exactly: 3,3,5,3,6. Good.

Need first paragraph not use numbering. It's a single p. Good.

Need maybe no <html> or body. Good.

Let's formulate final output.

Before final, let's double-check content "无死角视角" original says "正侧面双摄像头的高清直播与无死角视角" We must "与原文表达主旨保持高度关联与一致性". We include "正侧面双摄像头高清直播与无死角视角" or equivalent. Good.

Let's refine each sentence to be polished and natural, avoid overly marketing puffery? It's app store intro. Fine.

Need maybe "超强抓力让抓取成功率大幅提升" feature. We include.

Let's verify each point <=50 Chinese chars? Let's approximate. They said "每条约50字内" not strict? It's in task. We can keep concise. Let's count some:

Software intro:
p1 "1. 冒险达人面向喜爱抓娃娃与休闲游艺的用户，提供一站式线上娱乐体验。" Count excluding numbers/punct? Chinese characters: 冒险达人面向喜爱抓娃娃与休闲游艺的用户提供一站式线上娱乐体验 = maybe 29. Good.

p2 "2. 内置娃娃机、推币机等丰富玩法，覆盖多种线下经典游艺项目。" Characters: 内置娃娃机推币机等丰富玩法覆盖多种线下经典游艺项目 = 24. Good.

p3 "3. 用户可远程实时操控设备，用高清直播与多角度视野完成精准抓取。" Characters: 用户可远程实时操控设备用高清直播与多角度视野完成精准抓取 = 28. Good.

Operation:
p1 "打开应用后选择娃娃机或推币机，进入在线设备房间即可启动直播画面。" Characters: 打开应用后选择娃娃机或推币机进入在线设备房间即可启动直播画面 = 28. Good.
p2 "通过屏幕摇杆控制机器抓手，结合正侧面双摄像头调整角度并确认目标。" = 通过屏幕摇杆控制机器抓手结合正侧面双摄像头调整角度并确认目标 = 29.
p3 "点击抓取按钮实时操作，若成功抓中，可在线查看战利品与发货信息。" = 点击抓取按钮实时操作若成功抓中可在线查看战利品与发货信息 = 26.

Recommendation:
p1 "无需前往商场，在家即可随时畅玩娃娃机和推币机等热门游艺。" = 无需前往商场在家即可随时畅玩娃娃机和推币机等热门游艺 = 26.
p2 "双摄像头高清直播提供正侧面无死角视野，抓取位置看得更清楚。" = 双摄像头高清直播提供正侧面无死角视野抓取位置看得更清楚 = 28.
p3 "超强抓力提升抓取成功率，减少反复尝试带来的挫败感。" = 超强抓力提升抓取成功率减少反复尝试带来的挫败感 = 24.
p4 "多种玩法集中在一个应用，满足休闲娱乐和收集奖励的多元需求。" = 多种玩法集中在一个应用满足休闲娱乐和收集奖励的多元需求 = 26.
p5 "实时操控体验真实有趣，适合碎片时间放松和好友分享比拼。" = 实时操控体验真实有趣适合碎片时间放松和好友分享比拼 = 24.

Usage:
p1 "下载安装后在首次启动时同意摄像头与网络权限，确保直播操控流畅。" = 下载安装后在首次启动时同意摄像头与网络权限确保直播操控流畅 = 30.
p2 "注册并登录账号，充值或领取体验币后即可选择设备开始游玩。" = 注册并登录账号充值或领取体验币后即可选择设备开始游玩 = 26.
p3 "建议在稳定网络环境下使用，并根据画面提示调整手机方向和操作手势。" = 建议在稳定网络环境下使用并根据画面提示调整手机方向和操作手势 = 29.

Feature:
p1 "真实设备在线操控，手指轻点即可远程玩线下同款娃娃机。" = 真实设备在线操控手指轻点即可远程玩线下同款娃娃机 = 24.
p2 "正侧面双摄像头同步直播，抓取视角更全面，减少盲区误判。" = 正侧面双摄像头同步直播抓取视角更全面减少盲区误判 = 25.
p3 "高清画面配合实时回传，无需到店也能感受现场游玩氛围。" = 高清画面配合实时回传无需到店也能感受现场游玩氛围 = 25.
p4 "超强抓力设计有效提升出爪成功率，让奖励更容易到手。" = 超强抓力设计有效提升出爪成功率让奖励更容易到手 = 24.
p5 "娃娃机、推币机等多项玩法一站聚合，切换选择更轻松。" = 娃娃机推币机等多项玩法一站聚合切换选择更轻松 = 23.
p6 "随时打开应用即可开玩，通勤、居家等碎片场景都适合体验。" = 随时打开应用即可开玩通勤居家等碎片场景都适合体验 = 24.

All good.

Need maybe maybe "推币机" appears two times. Fine.

Let's think if the first p needs "冒险达人" exactly. Yes.

Potential issue: User requested "禁止 markdown 代码围栏、禁止任何解释或前言后记。" We must not output "```html" or comments. We'll output directly.

Potential issue: Since output should be "可直接写入数据库的 HTML 正文片段", no wrapper, newlines okay. We'll include newlines for readability, but they are part of fragment acceptable. No code block.

Now, should the non-测评 sections each have "若干" p, with each point "单独成行", meaning no blank line between p? In HTML, p elements are block, no blank line needed. We output one p per line. Good.

Should the sequence start exactly with a <p> and no whitespace before? We can start with `<p>`.

Potential issue: In "整体结构", "第一个元素是一个 <p>：依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍。" The first p should be an original rewrite of the "首段原文", including title. Our first p is good.

Need maybe mention "软件名称" not only first p? Already all h3 include. Good.

Need perhaps "首段" not "首段模板" with "标题：冒险达人" maybe title appears as text. Good.

Let's consider if the final evaluation must "全文不超过 200 字（含标点）" in the "测评" paragraph only? The instruction says "最后一个小节固定... 仅 1 个 <p>，不要序号与分条，一段短评，全文不超过 200 字（含标点）." "全文" likely refers to the测评 paragraph, not entire HTML. Wait "全文不超过 200 字" could mean entire generated excerpt? But since entire output has many paragraphs >200, impossible. Thus must refer to the测评段. The测评 task says "总字数不超过 200 字". So our测评 p only. Good.

Let's count final测评 p length to be safe. We can write a 120-160 char paragraph. Need no numbering "1." etc. Let's craft final p carefully:

"冒险达人把线下游艺厅常见的娃娃机、推币机搬到手机上，用双摄像头直播和远程操控解决了“看不见、抓不准”的痛点。对喜欢抓娃娃又没时间逛商场的人来说，它提供了一种低门槛的新选择。超强抓力提升了成就感，多种玩法也增加了留存乐趣；若后续在设备排队和发货体验上持续优化，会更值得一试。"

Count Chinese characters including punctuation: Let's count:
冒险达人把线下游艺厅常见的娃娃机、推币机搬到手机上，用双摄像头直播和远程操控解决了“看不见、抓不准”的痛点。对喜欢抓娃娃又没时间逛商场的人来说，它提供了一种低门槛的新选择。超强抓力提升了成就感，多种玩法也增加了留存乐趣；若后续在设备排队和发货体验上持续优化，会更值得一试。

Let's roughly count characters (Chinese + punctuation):
1 冒 2 险 3 达 4 人 5 把 6 线 7 下 8 游 9 艺 10 厅 11 常 12 见 13 的 14 娃 15 娃 16 机 17 、18 推 19 币 20 机 21 搬 22 到 23 手 24 机 25 上 26 ，27 用 28 双 29 摄 30 像 31 头 32 直 33 播 34 和 35 远 36 程 37 操 38 控 39 解 40 决 41 了 42 “43 看 44 不 45 见 46 、47 抓 48 不 49 准 50 ”51 的 52 痛 53 点54 。55 对56 喜57 欢58 抓59 娃60 娃61 又62 没63 时64 间65 逛66 商67 场68 的69 人70 来71 说72 ，73 它74 提75 供76 了77 一78 种79 低80 门81 槛82 的83 新84 选85 择86 。87 超88 强89 抓90 力91 提92 升93 了94 成95 就96 感97 ，98 多99 种100 玩101 法102 也103 增104 加105 了106 留107 存108 乐109 趣110 ；111 若112 后113 续114 在115 设116 备117 排118 队119 和120 发121 货122 体123 验124 上125 持126 续127 优128 化129 ，130 会131 更132 值133 得134 一135 试136 。137. Hmm 137 including punctuation. Great under 200.

But note use Chinese quotes “ ” are each punctuation? Counted as one each, fine. Good.

Need maybe not use "线上" in final review? Fine.

Now, but ensure the entire output no extra whitespace or newline before h3? Accept.

Let's think if there is a need to "仅按下列「小节顺序」输出" and "已从词池随机抽 3～6 个非测评小节" — maybe perhaps they only want a subset, but by listing all sections as tasks, maybe the system is asking us to simulate the selection? Let's revisit.

The user wrote:

【整体结构】顺序必须严格如下：
1) 第一个元素是一个 <p>：依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍。
2) 然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>（标题须与下列 {H3title} 完全一致），非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**（本次要求：用序号分条：每条单独成行，行首为「数字+英文句点+空格」（示例 1. 2. 3.），全篇保持同一格式）；**每节要点条数以该小节下方【条数】为准**（在 3～6 条之间且各节独立）。
3) **最后一个小节固定为「冒险达人测评」**（已在下列顺序末位）：**仅 1 个** <p>，**不要**序号与分条，**一段**短评，**全文不超过 200 字**（含标点）。

【小节顺序】软件介绍 → 操作指南 → 推荐理由 → 使用说明 → 特色 → 测评

Then all sections have detailed tasks. In the generation, they specify "已从词池随机抽 3～6 个非测评小节" but then they list all in order, not "随机选择". However, "【小节顺序】" maybe means the order from which to output selected ones. It includes "测评" last. But if random subset, they'd likely not provide tasks for all sections? They did. Could be automatic template with all possible tasks; the actual "随机抽" has already happened but omitted from user's visible prompt? No, the prompt says "已从词池随机抽 3～6 个非测评小节" but doesn't say which. It says "优先保留采集中有原文者；若含「介绍」「简介」类关键词会固定...". Here "软件介绍" includes "介绍", so it is fixed first. But which others are selected? The "小节顺序" lists all 5 non-测评. Maybe perhaps "从词池随机抽3~6个" means selecting among a "词池" of possible keywords, and this specific prompt has a "小节顺序" with all possible sections? Let's parse the text: "然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）"

Actually, "下列「小节顺序」" is "软件介绍 → 操作指南 → 推荐理由 → 使用说明 → 特色 → 测评" — six items. The phrase "已从词池随机抽 3～6 个非测评小节" could mean the entire "下列" list is what was randomly drawn from the pool, and "优先保留采集中有原文者" etc. It may be that all non-测评 sections listed are the selected ones (because 6 is within 3-6). They then say "全文共 4～7 个 h3" — 6 non-测评 + 测评 = 7, maximum. So it's valid to output all six. If they intended only a subset, they wouldn't require every section's "条数" perhaps. So all sections is safe.

Potential issue: "若含「介绍」「简介」类关键词会固定排在首段 <p> 之后的第一个 <h3>" — Since they have "软件介绍" with "介绍", it's fixed as first h3. Good.

Now, one more nuance: The first p must "自然出现" app name. Our first p starts with "冒险达人". Good.

Need maybe "应用商店的介绍文案" style: Should sound like app store product page. Our first paragraph fine.

Need maybe "禁止输出 <img>（应用截图由系统在生成后按小节顺序自动插入，模型无需处理图片）" We don't.

Need no markdown code fence. We won't.

Potential issue: The output must be "一段可直接写入数据库的 HTML 正文片段" Could "正文片段" include newlines? Yes. Not a problem.

Let's consider if "操作指南" with 3 points should perhaps not include "1. 2. 3." in each section reset; yes.

Potential issue: We need be "原创改写" not repeat original verbatim. Our first paragraph uses similar words but reorganized. Points are original.

Potential issue: "必须包含游戏/软件名称" maybe first p only. Good.

Let's see if all sections use "p" for each serial point. Yes.

Potential issue: "允许的 HTML 标签仅有：<p>、<h3>。" We use only p and h3. Good.

Potential issue: "不要输出 <html><body> 等外壳" We don't.

Now, maybe we should consider the evaluation of "全文不超过 200 字" — The "测评" part is ≤200; our full paragraph is 136. Good.

Let's perhaps refine final测评 to be more "拟人化" as required: "它把线下游艺厅常见的娃娃机、推币机搬到手机上..." already personifies "它". Good.

Maybe "可参考信息" from original: "带给用户前所未有的线上娱乐体验" Should we include "前所未有" maybe not necessary. The first paragraph can include "新奇的线上娱乐体验" perhaps. Our first paragraph: "冒险达人是一款集合娃娃机、推币机等丰富玩法的线上娱乐应用，让用户无需前往商场，即可通过手机实时操控真实设备；借助正侧面双摄像头高清直播与无死角视角，能精准抓取娃娃，超强抓力大幅提升抓取成功率。" This covers. Good.

Need maybe "在该软件内您无需再去商场" we say "无需前往商场". Fine.

Need "正侧面双摄像头" maybe "正侧面" as "正/侧面"? We use "正侧面" exactly. Good.

Need "无死角视角" we say "无死角视角". Good.

Need "实时操控娃娃机等设备" we say "实时操控真实设备". Good.

Now, let's think about output formatting with newlines. We'll produce:

<p>冒险达人是一款...</p>
<h3>冒险达人软件介绍</h3>
<p>1. ...</p>
...
<h3>冒险达人测评</h3>
<p>...</p>

No blank lines? Could include blank lines between blocks. In HTML, blank lines don't matter. But "禁止任何解释或前言后记" not about whitespace. Fine. To be safe, no extra blank lines? We'll include line breaks between elements, but no empty lines. A single newline after each tag is okay. This is not markdown.

Need maybe "首段 p" should not be inside a <p> with newline before? Fine.

Let's ensure no indentation with spaces? Fine.

Potential issue: "禁止 markdown 代码围栏" means no triple backticks. We won't.

Potential issue: "全文共 4～7 个 <h3>" If we output 7 h3, valid. But if some automator expects random subset, maybe output length not an issue. We must follow "全部小节" because each has tasks. Yes.

Let's double-check the section counts in output:

- 软件介绍: 3 p
- 操作指南: 3 p
- 推荐理由: 5 p
- 使用说明: 3 p
- 特色: 6 p
- 测评: 1 p

Total p after first = 21 plus first p = 22. h3 = 6. Wait h3 count: 6 non测评 + 1测评 = 6? Count: 软件介绍, 操作指南, 推荐理由, 使用说明, 特色, 测评 = 6 h3? Yes six! I earlier said 6 non-测评 + 测评 = 7, but the non-测评 list is 5, not 6? Let's re-evaluate: "软件介绍 → 操作指南 → 推荐理由 → 使用说明 → 特色" = five non-测评. Plus "测评" = 6 h3 total. So h3 count = 6, which is within 4-7. Good. Non-测评 = 5, within 3-6. Great.

Actually list has "软件介绍" (1), "操作指南"(2), "推荐理由"(3), "使用说明"(4), "特色"(5), "测评"(6). So h3=6. Fine.

Need maybe "末位固定测评" = h3 冒险达人测评. Good.

Now, let's perhaps check the first p length not constrained? No.

Let's consider if "应用名称（须在首段 <p> 正文中自然出现）" means the app name must appear in "首段" exactly maybe "冒险达人". We have.

Potential issue: "标题：冒险达人" might imply the first p should include "标题：冒险达人"? No, the first p should be intro; h3 headings include app names. We shouldn't output "标题：". Good.

Potential issue: They asked "不要输出 <html><body> 等外壳，不要 script/style." We obey.

Potential issue: "每个小节先 <h3>（标题须与下列 {H3title} 完全一致）" We use exact h3 text. Good.

Potential issue: "非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**" We use <p> per point. Good.

Potential issue: "每条单独成行，行首为「数字+英文句点+空格」（示例 1. 2. 3.）" We have "1. " with a space after period. In HTML, the space after "1." is in source, good. The instructions "数字+英文句点+空格" means e.g., "1. " exactly. We have.

Potential issue: "全篇保持同一格式" We consistently use "1. " etc. Good.

Potential issue: "每节要点条数以该小节下方【条数】为准" We have exact.

Potential issue: "本条数各节独立" We reset numbering each section. Good.

Potential issue: "每条约50字内" We are under.

Potential issue: "软件介绍" section "本节固定 3 条" We did.

Potential issue: "操作指南" "本节固定 3 条" We did.

Potential issue: "推荐理由" "本节固定 5 条" We did.

Potential issue: "使用说明" "本节固定 3 条" We did.

Potential issue: "特色" "本节固定 6 条" We did.

Potential issue: "测评" "仅 1 个 <p>, 不要序号" We did.

Now, could the system expect h3 heading text to include only "冒险达人" plus section keyword, e.g., "冒险达人软件介绍", exactly. Good.

Let's consider if "首段 p" should be "依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍。" They wrote "标题：冒险达人；原文首段：冒险达人是一款..." The first p must be an "original rewrite" of the original first paragraph, not "copy". Our first p is a rewrite but similar. Let's maybe make it more distinct while retaining all elements:

"冒险达人是一款聚合娃娃机、推币机等多样游艺玩法的线上娱乐应用。用户无需专程去商场，用手机就能实时操控真实设备；通过正侧面双摄像头高清直播与无死角视角，可清楚观察抓取位置，再配合超强抓力，让抓取成功率明显提升。"

This is more app-store style and reorganized. Contains app name and all key. I like this better than previous because "聚合" and "多样游艺玩法". Let's use this. Need around 110 chars? Fine. Let's check no "前所未有" but okay.

First p: "冒险达人是一款聚合娃娃机、推币机等多样游艺玩法的线上娱乐应用。用户无需专程去商场，用手机就能实时操控真实设备；通过正侧面双摄像头高清直播