six-pager

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

/six-pager -- Amazon-style Narrative Memo and PRFAQ

/six-pager -- 亚马逊风格叙事备忘录与PRFAQ

Generate decision documents in the format Amazon has used since June 9, 2004 — when Bezos banned PowerPoint and required every meeting to open with a written narrative.

This skill produces the document. It does not replace the discipline. The original Amazon practice takes a week (draft → review → set aside → edit → final). The skill compresses the structure but preserves the prose-quality bar.

生成采用亚马逊自2004年6月9日起使用格式的决策文档——当时贝索斯禁用了PowerPoint，要求每场会议都以书面叙事内容开场。

本工具负责生成文档，但无法替代撰写过程中的自律要求。亚马逊最初的实践流程需要一周时间（草稿→评审→搁置→编辑→终稿）。本工具压缩了流程结构，但保留了对写作质量的高标准。

Why This Works

为何此工具有效

Three independent traditions converge on a single principle: writing is the thinking instrument, not the documentation of it.

Bezos (Amazon, 2004): "There is no way to write a six-page narratively structured memo and not have clear thinking."
Strunk (1918, restated by E.B. White 1959): "Vigorous writing is concise. A sentence should contain no unnecessary words, a paragraph no unnecessary sentences, for the same reason that a drawing should have no unnecessary lines and a machine no unnecessary parts." (Rule 17)
Anthropic (Claude Code Best Practices): "For every line, ask: If I removed this, would Claude make a mistake? If not, remove it."

Same rule, three scales: line, sentence, instruction. This skill enforces all three.

三个独立的理念汇聚于同一原则：写作是思考的工具，而非思考结果的记录载体。

贝索斯（亚马逊，2004）： "撰写一份6页的叙事结构备忘录，不可能没有清晰的思路。"
斯特伦克（1918年，E.B.怀特于1959年重申）： "有力的写作是简洁的。句子不应包含多余的词，段落不应包含多余的句子，道理就像绘图不应有多余的线条、机器不应有多余的部件一样。"（规则17）
Anthropic（Claude代码最佳实践）： "对于每一行内容，问自己：如果我删掉这一行，Claude会出错吗？如果不会，就删掉它。"

同一规则，三个层面：行、句子、指令。本工具会同时强制执行这三个层面的要求。

When to Use This vs Sibling Skills

何时使用本工具及同类工具

Skill	Use For	Mode
`/six-pager memo`	Decisions, strategy, status, "should we?" with stakes	Narrative memo, 6 sections, hard 6-page cap
`/six-pager prfaq`	Product / feature / launch ideas where customer outcome is the question	Press release + 2 FAQs, work backwards from launch
a long-form writing/voice skill	Prose essays, narrative, sensory writing	Long-form drafting, not decision-doc structure
`/council-review`	Open question with multiple valid answers	Multi-agent deliberation, not document generation
`/adversarial-review`	Stress-test a finished artifact (including a six-pager!)	Single-critic attack mode

Natural pairing: Generate the document with

/six-pager

, then stress-test with

/adversarial-review

, then deliberate the recommendation with

/council-review

. That's the full Amazon-style decision pipeline.

工具	适用场景	模式
`/six-pager memo`	决策、战略、状态汇报、带有风险的“是否应该？”类问题	叙事备忘录，6个章节，严格限制6页篇幅
`/six-pager prfaq`	以客户成果为核心的产品/功能/发布创意	新闻稿+2份FAQ，从发布倒推撰写
长文写作/风格工具	散文、叙事、感官类写作	长文草稿撰写，不涉及决策文档结构
`/council-review`	存在多种合理答案的开放性问题	多Agent审议，不生成文档
`/adversarial-review`	对已完成的成果（包括6页备忘录！）进行压力测试	单一批评者攻击模式

最佳搭配流程： 使用

/six-pager

生成文档，再用

/adversarial-review

进行压力测试，最后用

/council-review

审议建议。这就是完整的亚马逊风格决策流程。

Arguments

参数

Argument 1 (required): The topic, draft, or file path. One of:
- A short topic statement:
```
"Should we migrate the API to cursor-based pagination?"
```
- A path to an existing draft:
```
docs/pagination-decision-draft.md
```
- A pasted block of context
Mode flag (optional):
```
memo
```
(default for decisions) or
```
prfaq
```
(for launches). If omitted, the skill auto-classifies and asks to confirm.
--silent-read
: Adds a 20-30 minute "silent read" simulation to the output — produces an annotated version with margin questions a senior reviewer would write.
--strunk-only
: Skips structural generation; treats the input as an existing draft and runs only the Strunk + Anthropic prose audit. Useful for cleaning up an in-flight document.

参数1（必填）： 主题、草稿或文件路径。可选类型：
- 简短主题陈述：
```
"我们是否应将API迁移为基于游标分页？"
```
- 现有草稿的文件路径：
```
docs/pagination-decision-draft.md
```
- 粘贴的上下文内容块
模式标记（可选）：
```
memo
```
（决策场景默认模式）或
```
prfaq
```
（发布场景）。若省略，工具会自动分类并请求确认。
--silent-read
：在输出中添加20-30分钟的“静默阅读”模拟——生成带有资深评审者会标注的页边问题的注释版文档。
--strunk-only
：跳过结构生成；将输入视为现有草稿，仅执行Strunk+Anthropic写作规范审核。适用于优化正在撰写中的文档。

Procedure

流程

Phase 1: SCOPE

阶段1：范围确认

Parse flags from
```
$ARGUMENTS
```
.
Classify the input:
- Topic — short statement, no draft. Generate from scratch.
- File path — existing draft. Read it; treat as starting point.
- Pasted text — treat as starting context.
If mode flag is missing:
- Auto-classify: input contains "launch" / "ship" / "announce" / "new product" / "release" → suggest
```
prfaq
```
  . Otherwise → suggest
```
memo
```
  .
- Confirm with the user before proceeding.
Confirm scope back to the user in one sentence: "Producing [memo|prfaq] for [topic]. [Optional: --silent-read enabled.]"

解析
```
$ARGUMENTS
```
中的标记。
对输入进行分类：
- 主题 —— 简短陈述，无草稿。从头开始生成文档。
- 文件路径 —— 现有草稿。读取文件内容，作为生成起点。
- 粘贴文本 —— 作为初始上下文处理。
若未指定模式标记：
- 自动分类：输入包含“launch”/“ship”/“announce”/“new product”/“release”→建议使用
```
prfaq
```
  模式。否则→建议使用
```
memo
```
  模式。
- 继续执行前需获得用户确认。
用一句话向用户确认范围：“将为[主题]生成[memo/prfaq]文档。[可选：已启用--silent-read。]”

Phase 2: DRAFT

阶段2：草稿生成

Mode:

memo

(6-pager)

模式：

memo

（6页备忘录）

Generate the six canonical sections. Each section has a specific purpose and a strict prose discipline:

Introduction (~0.5 page) — What is this document? Who is the audience? What is the ask? Frame the decision in one paragraph.
Goals (~0.5 page) — Specific, measurable outcomes. Each goal must be verifiable. Numbers required where possible.
Tenets (~0.5 page) — The principles this proposal optimizes for. No qualifiers (per Strunk's "avoid leeches" reminder). Tenets are unhedged commitments. Format:
```
1. [Tenet name]: [single declarative sentence].
```
State of the Business (~1.5 pages) — Current data. Honest. Concrete numbers, named systems, dated facts. Apply Strunk Rule 16 (definite, specific, concrete) ruthlessly.
Lessons Learned (~1 page) — What did past attempts teach? What is different this time? Each lesson cites a specific past event or experiment, not abstract principles.
Strategic Priorities (~2 pages) — The recommended action and rationale. Specific milestones, owners, timelines.
Appendix (unlimited) — Visuals, supporting data, alternative analyses considered and rejected (with reasons).

(V2) Premortem (required in Strategic Priorities or the Appendix). Assume it is 12 months later and this decision failed. Write the 3 most likely causes, working backward from the failure. This is forward-looking risk-finding that "Lessons Learned" (backward-looking) structurally misses — Kahneman's single highest-value decision technique, and it pairs with the verification: each premortem cause should map to a thing you'll monitor. Not generic "risks" — concrete failure scenarios.

Hard constraint: sections 1-6 combined must fit in 6 pages of 11pt body text. The skill renders a page count and warns if over.

生成六个标准章节。每个章节都有特定用途和严格的写作规范：

引言（约0.5页） —— 本文档是什么？受众是谁？核心诉求是什么？用一段文字阐述决策背景。
目标（约0.5页） —— 具体、可衡量的成果。每个目标必须可验证，尽可能包含数字。
原则（约0.5页） —— 本提案优化所遵循的准则。禁止使用限定词（遵循Strunk的“避免冗余修饰”提示）。原则是明确的承诺。格式：
```
1. [原则名称]：[单一陈述句]。
```
业务现状（约1.5页） —— 当前数据。内容需真实，包含具体数字、命名系统、日期事实。严格执行Strunk规则16（明确、具体、具象）。
经验教训（约1页） —— 过往尝试带来了什么启示？本次有何不同？每条教训需引用具体的过往事件或实验，而非抽象原则。
战略优先级（约2页） —— 推荐行动方案及理由。包含具体里程碑、负责人、时间线。
附录（无限制） —— 可视化内容、支持数据、已考虑并否决的替代分析（需说明理由）。

（V2版本新增）事前复盘（必填，置于战略优先级或附录中）。假设12个月后该决策失败，写出3个最可能的原因，从失败结果倒推分析。这是前瞻性的风险排查，弥补了“经验教训”（回顾性）在结构上的不足——这是卡尼曼提出的最高价值决策技巧，且需与验证机制结合：每个事前复盘的原因都应对应一个你将监控的指标。不能是笼统的“风险”，而需是具体的失败场景。

硬性约束： 第1-6章节总篇幅必须适配11号字体的6页内容。工具会计算页数并在超限时发出警告。

Mode:

prfaq

(Press Release + FAQs)

模式：

prfaq

（新闻稿+FAQ）

Generate three sub-documents. Work backwards — write the press release first, as if the product already shipped.

Press Release (~1 page) — Customer-facing language. Benefit-led. Dated for the (hypothetical) launch day. Format follows Amazon's internal PR conventions:
- Headline (one line, customer-outcome-oriented)
- Sub-headline (one line, who-it's-for + key-benefit)
- Lede paragraph (city, date, "Today X announced Y...")
- Problem paragraph (what was hard before this product existed)
- Solution paragraph (how the product solves the problem)
- Customer quote (fictional but realistic)
- Internal quote (from the team, on why this matters)
- Call to action / availability
External FAQ (~1 page) — Questions a journalist or customer would ask:
- What is this?
- How is it different from [obvious alternative]?
- How much does it cost?
- When can I get it?
- What does it not do?
Internal FAQ (~1-2 pages) — Questions Amazon-style leadership would ask:
- Why now?
- What does success look like? (specific metrics, specific dates)
- What are the dependencies?
- What is the failure mode?
- What are we choosing not to build?
- What would cause us to kill this?

The PRFAQ test: if you can't write a credible press release, the product doesn't exist yet — not in the form that matters.

生成三个子文档。倒推撰写——先写新闻稿，假设产品已发布。

新闻稿（约1页） —— 面向客户的表述，以利益为导向。标注（假设的）发布日期。格式遵循亚马逊内部PR规范：
- 标题（一行，以客户成果为核心）
- 副标题（一行，说明受众+核心利益）
- 导语段落（城市、日期、“今日X宣布Y...”）
- 问题段落（该产品出现前用户面临的痛点）
- 解决方案段落（产品如何解决问题）
- 客户引用（虚构但符合现实）
- 内部团队引用（说明该产品的重要性）
- 行动号召/可用性说明
外部FAQ（约1页） —— 记者或客户可能提出的问题：
- 这是什么？
- 它与[明显竞品]有何不同？
- 价格是多少？
- 何时可用？
- 它不具备哪些功能？
内部FAQ（约1-2页） —— 亚马逊风格领导层可能提出的问题：
- 为什么是现在？
- 成功的标准是什么？（具体指标、具体日期）
- 依赖项有哪些？
- 失败模式是什么？
- 我们选择不开发什么？
- 什么情况会导致我们终止该项目？

PRFAQ测试标准： 如果你无法写出可信的新闻稿，说明产品尚未成型——至少不是以有意义的形式存在。

Phase 3: CONSTRAIN

阶段3：篇幅约束

After Phase 2 produces a draft, measure it:

bash

undefined

阶段2生成草稿后，计算篇幅：

bash

undefined

rough page count: assume 450 words/page at 11pt

粗略页数计算：假设11号字体每页450词

WORDS=$(wc -w < draft.md) PAGES=$(echo "scale=1; $WORDS / 450" | bc)


If over budget:
- For `memo`: identify the longest section, propose 30pct cuts that preserve load-bearing content
- For `prfaq`: identify whichever FAQ sub-section ran long; propose merging or trimming questions

Never silently truncate. Always show the user what's being proposed for cuts and why.

WORDS=$(wc -w < draft.md) PAGES=$(echo "scale=1; $WORDS / 450" | bc)


若超出篇幅限制：
- 对于`memo`模式：找出最长的章节，提出保留核心内容的30%删减建议
- 对于`prfaq`模式：找出过长的FAQ子章节，提出合并或删减问题的建议

绝不自动截断内容。必须向用户展示拟删减的内容及理由。

Phase 4: AUDIT (Strunk + Anthropic)

阶段4：审核（Strunk+Anthropic）

Run the prose-quality audit on the draft. Each finding cites the rule:

Check	Rule	Action
Passive voice	Strunk R14	Flag every passive construction; suggest active rewrite
Vague language	Strunk R16	Flag every "improvements," "various," "some," "many" without a number; demand specificity
Needless words	Strunk R17 + Anthropic removability	Flag wordy phrases ("in order to" → "to"; "due to the fact that" → "because")
Qualifiers	E.B. White's "leeches" reminder	Flag every "rather," "very," "little," "pretty," "quite," "somewhat" — propose deletion
Parallel construction	Strunk R19	Flag inconsistent grammatical forms in lists / tenets / goals
Sentence-length monotony	Strunk R18	Flag stretches of 4+ sentences within 5 words of each other; propose variation
Topic sentence	Strunk R13	Flag any paragraph that buries its claim under setup. Lead each paragraph with its point — one paragraph = one step in the argument. The core narrative-memo discipline
Overstatement	E.B. White, do not overstate	Flag carefree superlatives and unhedged hype ("massive," "revolutionary," "huge," "game-changing"). A single one puts the reader on guard and taints the rest — demand the number or cut
Tenets with hedges	Bezos canon	Tenets are commitments. "We try to optimize for X" → "We optimize for X" or cut
Removability test	Anthropic	For each paragraph: would removing it cause the reader to make a wrong decision? If not, cut.

Produce a structured audit report:

undefined

对草稿进行写作质量审核。每项发现需标注对应的规则：

检查项	规则	行动
被动语态	Strunk规则14	标记所有被动结构；建议改写为主动语态
模糊表述	Strunk规则16	标记所有无数字支撑的“改进”“各种”“一些”“许多”；要求具体化
冗余词汇	Strunk规则17 + Anthropic可移除性准则	标记冗余短语（如“in order to”改为“to”；“due to the fact that”改为“because”）
限定词	E.B.怀特的“冗余修饰”提示	标记所有“rather”“very”“little”“pretty”“quite”“somewhat”——建议删除
平行结构	Strunk规则19	标记列表/原则/目标中不一致的语法结构
句子长度单调	Strunk规则18	标记连续4句以上长度相差不超过5词的段落；建议调整句式变化
主题句	Strunk规则13	标记任何将核心观点隐藏在铺垫内容后的段落。每段需以核心观点开头——一段对应一个论证步骤。这是叙事备忘录的核心规范
夸大表述	E.B.怀特“勿夸大”原则	标记随意使用的最高级和无依据的炒作词汇（如“massive”“revolutionary”“huge”“game-changing”）。哪怕只用一个这类词汇，都会让读者产生警惕并影响整体可信度——要求提供具体数字或删除该词汇
带模糊修饰的原则	贝索斯规范	原则是明确的承诺。“We try to optimize for X”需改为“We optimize for X”或删除该原则
可移除性测试	Anthropic准则	对于每个段落：删除它会导致读者做出错误决策吗？如果不会，就删除该段落。

生成结构化审核报告：

undefined

Prose Audit (Strunk + Anthropic)

写作规范审核（Strunk+Anthropic）

Findings: N critical, M important, K nit

发现： N项严重问题，M项重要问题，K项细微问题

Critical (block the document)

严重问题（阻碍文档通过）

Line 23: passive voice in Strategic Priorities ("It was decided that...") → "We decided to..."
Line 47: vague Goal ("improve performance") → demand a number
Tenet 3 hedges ("We strive to be customer-obsessed") → cut "strive to" or cut the tenet

第23行：战略优先级部分使用被动语态（"It was decided that..."）→ 改为"We decided to..."
第47行：目标表述模糊（"improve performance"）→ 要求提供具体数字
原则3存在模糊修饰（"We strive to be customer-obsessed"）→ 删除"strive to"或删除该原则

Important (fix in next pass)

重要问题（需在下一版本修复）

4 sentences in a row of 18-22 words in State of Business → vary
Parallel-form break in Goals: 3 noun-phrase goals + 1 verb-phrase goal
State of Business ¶2 buries its claim in the final sentence → lead the paragraph with it
Overstatement in Introduction ("a massive, game-changing opportunity") → name the number or cut

业务现状部分连续4句长度为18-22词→ 建议调整句式
目标部分平行结构不一致：3个名词短语目标+1个动词短语目标
业务现状第2段将核心观点放在最后一句→ 建议将核心观点放在段首
引言部分存在夸大表述（"a massive, game-changing opportunity"）→ 提供具体数字或删除该表述

Nit

细微问题

"in order to" appears 3 times → "to"
"various improvements" → name them or cut

undefined

"in order to"出现3次→ 改为"to"
"various improvements"→ 具体说明或删除

undefined

Phase 5: PRESENT

阶段5：呈现结果

Show the user:

Final document (or draft with audit annotations if
```
--silent-read
```
)
Page count vs budget
Audit summary (count of critical/important/nit findings)
Side-by-side: original draft vs proposed rewrites for any audit findings

Ask: "Apply the audit fixes? (yes / partial / no)"

If yes: apply fixes in place (surgical Edits if input was a file; show new full document otherwise). If partial: walk through each fix individually. If no: deliver the unaudited draft with a note that audit was skipped.

向用户展示：

最终文档（若启用
```
--silent-read
```
则展示带审核注释的草稿）
实际页数与限定页数对比
审核摘要（严重/重要/细微问题数量）
对比展示：原始草稿与审核建议的改写内容（针对所有审核发现）

询问用户：“是否应用审核修复建议？（是/部分/否）”

若选择“是”：直接应用修复（若输入为文件则进行精准编辑；否则展示完整的新文档）。若选择“部分”：逐一确认每项修复建议。若选择“否”：交付未审核的草稿，并标注已跳过审核步骤。

Phase 6: SAVE

阶段6：保存文档

Default save path: alongside the input file (if file input) or

<cwd>/six-pager-<topic-slug>-<YYYY-MM-DD>.md

--silent-read

was used, also produce

<same-name>-annotated.md

with margin questions in HTML comments or callout blocks.

默认保存路径：与输入文件同目录（若输入为文件）或

<当前工作目录>/six-pager-<主题缩写>-<YYYY-MM-DD>.md

。

若启用

--silent-read

，还需生成

<同名文件>-annotated.md

，将页边问题放在HTML注释或提示块中。

The Silent-Read Simulation (

--silent-read

)

静默阅读模拟（

--silent-read

）

The original Amazon meeting protocol: distribute the document at the start of the meeting (NOT pre-read), 20-30 minutes of silent reading with pens out, then discussion anchored to specific page/line annotations.

When

--silent-read

is enabled, after generating the document the skill simulates the silent read by producing a second pass:

For each section, generate 2-4 questions a senior reviewer would write in the margin. Examples:

"How confident is the 23pct number? Is this measured or estimated?"
"Tenet 2 contradicts Tenet 4 — which wins when they conflict?"
"Lesson Learned #3 says the prior attempt failed because of X. What evidence?"
"Strategic Priority #1 has no owner named."

This catches the "writer thinks it's clear; reader doesn't" failure mode before the actual meeting.

亚马逊原始会议流程：会议开始时分发文档（而非提前阅读），预留20-30分钟静默阅读时间（参会者可做笔记），随后基于具体页码/行的注释展开讨论。

启用

--silent-read

后，生成文档后工具会进行二次处理，模拟静默阅读环节：

针对每个章节，生成2-4个资深评审者会写在页边的问题。示例：

“23%这个数字的可信度如何？是实测值还是估算值？”
“原则2与原则4存在矛盾——冲突时以哪项为准？”
“经验教训3提到之前的尝试因X失败，有何证据支撑？”
“战略优先级1未指定负责人。”

这可以在实际会议前发现“作者认为表述清晰，但读者无法理解”的问题。

Quality Bar

质量标准

A six-pager passes if every check is true. Otherwise iterate.

Page count ≤ 6 (memo) or correct sub-document length (prfaq)
Every Goal has a number or specific verifiable outcome
Every Tenet is a declarative commitment with no qualifiers
State of Business cites at least 5 specific numbers / named systems / dated facts per page
Strategic Priorities name at least one owner per priority
Strunk audit: zero critical findings (passive voice, vague Goals, hedged Tenets); important findings acceptable with rationale
The "what would cause us to kill this?" question (PRFAQ Internal FAQ) has a real, specific answer — not "if performance suffers"

一份6页备忘录需满足所有以下检查项才算合格，否则需迭代优化。

页数≤6（memo模式）或符合子文档长度要求（prfaq模式）
每个目标都有数字或具体可验证的成果
每个原则都是无修饰的明确承诺
业务现状部分每页至少引用5个具体数字/命名系统/日期事实
战略优先级部分每个优先级至少指定一位负责人
Strunk审核：无严重问题（被动语态、模糊目标、带修饰的原则）；重要问题需有合理理由方可接受
PRFAQ内部FAQ中的“什么情况会导致我们终止该项目？”需有真实、具体的答案——不能是“如果性能下降”这类笼统表述

Gotchas

注意事项

Do not exceed 6 pages. The constraint IS the value. If it doesn't fit, the thinking isn't done.
Do not pad State of Business with backstory. Current data only. History belongs in Lessons Learned.
Do not hedge Tenets. A Tenet with "try to" or "strive to" is not a Tenet.
Do not generate fake numbers. If the Goal needs a metric and the user hasn't supplied one, ask. Generic placeholders ("X%") are not acceptable.
Do not confuse Goals with Tenets. Goals are outcomes (measurable). Tenets are principles (commitments). Both are needed; conflating them muddles thinking.
Do not skip the audit. Phase 4 is not optional. The whole point is prose discipline plus structure.
Do not auto-save. Phase 6 confirms with the user where to write.
Do not invoke this for prose articles. Use a dedicated long-form writing/voice skill instead — this skill is for decision docs, not essays.

请勿超过6页篇幅。 约束本身就是价值所在。如果内容无法压缩，说明思考尚未成熟。
请勿在业务现状部分添加背景故事。 仅包含当前数据。历史内容应放在经验教训部分。
请勿对原则添加模糊修饰。 带有“try to”或“strive to”的内容不能称为原则。
请勿生成虚假数字。 如果目标需要指标但用户未提供，需询问用户。不允许使用通用占位符（如“X%”）。
请勿混淆目标与原则。 目标是成果（可衡量）。原则是准则（承诺）。两者缺一不可；混淆会导致思路混乱。
请勿跳过审核步骤。 阶段4是必选项。本工具的核心价值就是写作规范加结构化流程。
请勿自动保存。 阶段6需获得用户确认后再保存。
请勿将本工具用于散文写作。 请使用专门的长文写作/风格工具——本工具仅用于决策文档，而非散文。

Changelog

更新日志

V2 (2026-05-27)

V2（2026-05-27）

Optimized via

skillforge optimize

(outcome research: decision-memo efficacy + decision science). six-pager was already strong; this is a focused, non-duplicative add:

Required premortem in memo mode — forward-looking "assume it failed, why?" failure analysis, each cause tied to a verification signal. Closes the gap that "Lessons Learned" (backward-looking) leaves open. Kahneman's highest-value decision technique; same decision-science line that informed council-review V2.
Outcome target: memos that surface the killer risk before the decision, not after. Sources: Kahneman (premortem); Amazon narrative practice (silent read, customer-back); grounded in the same decision-science line (premortem, mediating assessments).

通过

skillforge optimize

优化（基于决策备忘录有效性+决策科学的成果研究）。six-pager工具原本已具备较强能力，本次更新为针对性的非重复新增功能：

memo模式强制要求事前复盘——前瞻性的“假设失败，原因何在？”失败分析，每个原因都对应一个验证信号。弥补了“经验教训”（回顾性）的不足。这是卡尼曼提出的最高价值决策技巧；与council-review V2版本基于同一决策科学逻辑。
目标成果：在决策前而非之后发现致命风险。参考资料：卡尼曼（事前复盘）；亚马逊叙事实践（静默阅读、以客户为核心）；基于同一决策科学逻辑（事前复盘、中介评估）。

References

参考资料

Bezos / Amazon canon:

Bryar, Colin & Carr, Bill — Working Backwards: Insights, Stories, and Secrets from Inside Amazon (St. Martin's Press, 2021) — canonical book
Slab — How Jeff Bezos Turned Narrative into Amazon's Competitive Advantage
Maestra — How to Write an Amazon 6-Pager (with Template)
Anecdote — What might Amazon's 6 page narrative structure look like?

Strunk & White canon:

Strunk, William Jr. & White, E.B. — The Elements of Style, 4th edition (Pearson Allyn & Bacon, 2000)
Public-domain text of Strunk's 1918 original: Bartleby — The Elements of Style

Anthropic prose-discipline:

Anthropic — Claude Code Best Practices — the removability test

Cognitive-style background:

Tufte, Edward — The Cognitive Style of PowerPoint (2003) — the case against bullet-points

贝索斯/亚马逊规范：

Colin Bryar & Bill Carr — Working Backwards: Insights, Stories, and Secrets from Inside Amazon（圣马丁出版社，2021）——权威书籍
Slab — How Jeff Bezos Turned Narrative into Amazon's Competitive Advantage
Maestra — How to Write an Amazon 6-Pager (with Template)
Anecdote — What might Amazon's 6 page narrative structure look like?

斯特伦克&怀特规范：

William Jr. Strunk & E.B. White — The Elements of Style，第4版（Pearson Allyn & Bacon，2000）
斯特伦克1918年原版的公有领域文本：Bartleby — The Elements of Style

Anthropic写作规范：

Anthropic — Claude Code Best Practices——可移除性测试

认知风格背景：

Edward Tufte — The Cognitive Style of PowerPoint（2003）——反对使用项目符号的论据

Testing

测试

This is a prompt-only skill. The shipped

tests/eval.sh

asserts the structural contract. To verify behavior end-to-end:

Pick a real decision:

"/six-pager Should we migrate the API to cursor-based pagination?"

(memo mode)

Verify all 6 sections generated, page count ≤ 6, audit run with findings

Pick a launch idea:

"/six-pager prfaq Acme AI assistant public launch"

(PRFAQ mode)

Verify press release + 2 FAQs generated, "what would cause us to kill this?" has a real answer
Run
```
--strunk-only
```
on an existing draft: verify only the audit fires, no new structure generated
Run
```
--silent-read
```
: verify margin questions are produced for each section

本工具仅基于提示词运行。已发布的

tests/eval.sh

用于验证结构约定。要进行端到端行为验证：

选择一个真实决策：

"/six-pager Should we migrate the API to cursor-based pagination?"

（memo模式）

验证生成所有6个章节，页数≤6，且已执行审核并生成发现

选择一个发布创意：

"/six-pager prfaq Acme AI assistant public launch"

（PRFAQ模式）

验证生成新闻稿+2份FAQ，“什么情况会导致我们终止该项目？”有真实答案
对现有草稿运行
```
--strunk-only
```
：验证仅执行审核，未生成新结构
运行
```
--silent-read
```
：验证每个章节都生成了页边问题