Back to Details

training-ai-prompt-writing

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

AI Prompt Writing for Marketing Teams — Training Guide

面向营销团队的AI提示词写作培训指南

<!-- dual-compat:start -->
<!-- dual-compat:start -->

Use when

适用场景

  • Produces a practical training guide for client teams on prompt engineering for marketing tasks — covering the Alpha-Beta-Gamma-Delta-Epsilon prompt structure, 10 prompt components, 5 prompting approaches, and 7 copywriting frameworks with worked East African examples. Invoke when the user says "create a prompt writing training guide", "teach my team how to use AI for marketing", "write a prompt engineering workshop", "AI copywriting training for staff", or needs a structured training document for client employees who use AI tools to produce marketing content.
  • Use this skill when it is the closest match to the requested deliverable or workflow.
  • 为客户团队生成面向营销任务的提示工程实用培训指南——涵盖Alpha-Beta-Gamma-Delta-Epsilon提示结构、10个提示组件、5种提示方法，以及带有东非实操案例的7个文案框架。当用户提出“创建提示词写作培训指南”“教我的团队如何用AI做营销”“撰写提示工程工作坊资料”“为员工提供AI文案写作培训”，或需要为使用AI工具制作营销内容的客户员工提供结构化培训文档时，可调用本技能。
  • 当本技能与用户要求的交付成果或工作流程最匹配时使用。

Do not use when

不适用场景

  • Do not use this skill for graphic design, video production, software development, or legal advice beyond the repository's stated scope.
  • Do not use it when another skill in this repository is clearly more specific to the requested deliverable.
  • 请勿将本技能用于图形设计、视频制作、软件开发或超出知识库规定范围的法律咨询。
  • 当知识库中其他技能明显更适合用户要求的交付成果时，请勿使用本技能。

Workflow

工作流程

  1. Collect the required inputs or source material before drafting, unless this skill explicitly generates the intake itself.
  2. Follow the section order and decision rules in this 
    SKILL.md
    ; do not skip mandatory steps or required fields.
  3. Review the draft against the quality criteria, then deliver the final output in markdown unless the skill specifies another format.
  1. 在起草前收集所需输入或素材，除非本技能明确说明可自行生成输入内容。
  2. 遵循本
    SKILL.md
    中的章节顺序和决策规则；请勿跳过必填步骤或必填字段。
  3. 根据质量标准审核草稿，然后以Markdown格式交付最终输出，除非技能指定其他格式。

Anti-Patterns

反模式

  • Do not invent client facts, performance data, budgets, or approvals that were not provided or clearly inferred from evidence.
  • Do not skip required inputs, mandatory sections, or quality checks just to make the output shorter.
  • Do not drift into out-of-scope work such as code implementation, design production, or unsupported legal conclusions.
  • 请勿编造未提供或无法从证据中明确推断的客户事实、绩效数据、预算或审批信息。
  • 请勿为缩短输出内容而跳过必填输入、必填章节或质量检查。
  • 请勿偏离范围开展工作，如代码实现、设计制作或无依据的法律结论。

Outputs

输出成果

  • A structured markdown document, plan, playbook, or strategy ready for client-facing or internal use.
  • 结构化的Markdown文档、计划、手册或策略，可直接用于客户对接或内部使用。

References

参考资料

  • Use the inline instructions in this skill now. If a 
    references/
    directory is added later, treat its files as the deeper source material and keep this 
    SKILL.md
    execution-focused.
<!-- dual-compat:end -->
  • 目前使用本技能中的内嵌说明。若后续添加
    references/
    目录，将其文件视为深层素材来源，本
    SKILL.md
    仍以执行为核心。
<!-- dual-compat:end -->

How to Use This Skill

如何使用本技能

Collect the Required Input below. Then generate the full training guide in four modules, substituting all bracketed placeholders with the client's specific details. Output is a complete, facilitator-ready training document — not a slide deck. For a slide deck version, use the 
deck-
prefix convention and build slides separately.
收集以下必填输入，然后生成包含四个模块的完整培训指南，将所有占位符替换为客户的具体信息。输出为适合主持人使用的完整培训文档——而非幻灯片。如需幻灯片版本，请使用
deck-
前缀约定并单独制作幻灯片。

Required Input

必填输入

Ask for the following before generating the training guide:
  • Client business name and industry — trading name and sector (e.g. telecoms, microfinance, FMCG, hospitality)
  • Country / city — default Uganda / East Africa
  • Primary goal — what the client wants the team to achieve after training
  • Team size and prior AI experience level — number of participants and experience (none / basic / intermediate)
  • Primary content types produced — captions, emails, blogs, WhatsApp broadcasts, SMS, ad copy, etc.
  • Preferred AI tools — ChatGPT, Gemini, Claude, or other (specify versions where known)
  • Training format — in-person half-day / virtual session / self-guided handout
生成培训指南前请获取以下信息：
  • 客户企业名称及行业——商号和领域（如电信、小额信贷、快速消费品、酒店业）
  • 国家/城市——默认乌干达/东非
  • 主要目标——客户希望团队在培训后达成的成果
  • 团队规模及AI使用经验水平——参与者人数和经验（无/基础/中级）
  • 主要制作的内容类型——标题文案、邮件、博客、WhatsApp广播、短信、广告文案等
  • 首选AI工具——ChatGPT、Gemini、Claude或其他工具（已知版本请注明）
  • 培训形式——线下半天/线上会议/自学手册

Output: Complete Training Guide

输出：完整培训指南

Generate the following four modules in full. Use the client's name, industry, preferred AI tools, and primary content types throughout. Write in plain English — no jargon. Tone: practical, encouraging, professional.
完整生成以下四个模块。全程使用客户名称、行业、首选AI工具和主要内容类型。语言通俗易懂——无专业术语。语气：实用、鼓励、专业。

Training Overview

培训概述

Programme: AI Prompt Writing for Marketing Teams Total Duration: Approximately 2.5 hours (150 minutes) Audience: Marketing, communications, and content staff with any level of AI experience Format: [Insert training format] Prepared for: [Client Business Name] Industry: [Industry] Primary Sources: Upadhyay, M.A. (2024) Generative AI for Marketing (Packt); Anderson, D. (2022) AI in Digital Marketing Training Guide (Self-published)
项目： 面向营销团队的AI提示词写作培训 总时长： 约2.5小时（150分钟） 受众： 具备任何AI使用经验的营销、传播及内容工作人员 形式： [填入培训形式] 为以下对象准备： [客户企业名称] 行业： [行业] 主要参考资料： Upadhyay, M.A. (2024) Generative AI for Marketing (Packt); Anderson, D. (2022) AI in Digital Marketing Training Guide (Self-published)

Module 1: Why Prompt Quality Matters (30 minutes)

模块1：提示词质量的重要性（30分钟）

Learning Objective

学习目标

Explain why the quality of AI output is directly determined by the quality of the prompt — and establish the principle that good prompts are a learnable, professional skill.
解释AI输出质量直接取决于提示词质量的原因——并确立优质提示词是可习得的专业技能这一原则。

1.1 What AI Can and Cannot Do

1.1 AI能做什么，不能做什么

Explain the following to participants:
AI tools such as ChatGPT, Gemini, and Claude are language prediction engines. They produce statistically probable text based on instructions. They do not:
  • Know your brand unless you tell them
  • Understand your audience unless you describe them
  • Know what to avoid unless you specify constraints
  • Default to quality — they default to plausible
AI tools excel at:
  • Drafting and editing at speed
  • Generating multiple variations quickly
  • Applying a named copywriting framework consistently
  • Adapting tone, length, and format on instruction
向参与者解释：
ChatGPT、Gemini和Claude等AI工具是语言预测引擎。它们根据指令生成统计上合理的文本。它们无法：
  • 知晓你的品牌，除非你告知它们
  • 理解你的受众，除非你描述清楚
  • 知道需要规避的内容，除非你明确限制
  • 默认生成高质量内容——它们默认生成看似合理的内容
AI工具擅长：
  • 快速起草和编辑内容
  • 快速生成多个版本
  • 一致应用指定的文案框架
  • 根据指令调整语气、篇幅和格式

1.2 The GIGO Principle

1.2 GIGO原则

Apply the GIGO principle (Garbage In, Garbage Out — Anderson, 2022) as the session's opening argument: vague instructions produce vague output. Demonstrate this directly.
Live demonstration — run this exercise:
Ask the AI tool: "Write me a caption."
Show the output to participants. Ask: "Would you post this?"
Then ask: "You are a social media copywriter for a Ugandan mobile money brand targeting market traders aged 30–50 in Kampala. Write one Instagram caption using the PAS framework (Problem → Agitate → Solution) for a new zero-fee transfer promotion. Under 150 characters. End with a call to action. No exclamation marks."
Show the second output. Discuss the quality gap.
将GIGO原则（垃圾进，垃圾出——Anderson, 2022）作为开场论点：模糊的指令会产生模糊的输出。直接进行演示。
现场演示——开展以下练习：
向AI工具提问：“给我写一条标题文案。”
向参与者展示输出内容。提问：“你会发布这条内容吗？”
然后提问：“你是一名面向乌干达移动支付品牌的社交媒体文案师，目标受众是坎帕拉30-50岁的市场商贩。使用PAS框架（问题→激化→解决方案）为新的零手续费转账活动撰写一条Instagram标题文案。字数不超过150字符。结尾包含行动号召。禁止使用感叹号。”
展示第二次的输出内容，讨论两者的质量差距。

1.3 The Quality Gap

1.3 质量差距

Define the quality gap as the difference between raw AI output and human-quality output ready for publishing. The gap is closed by:
  1. Providing complete context (Beta element)
  2. Assigning an expert role (Alpha element)
  3. Specifying the exact task and format (Gamma and Epsilon elements)
  4. Restricting what to avoid (Delta element)
将质量差距定义为原始AI输出与可直接发布的人工质量输出之间的差异。可通过以下方式缩小差距：
  1. 提供完整背景信息（Beta元素）
  2. 指定专家角色（Alpha元素）
  3. 明确具体任务和格式（Gamma和Epsilon元素）
  4. 限制需规避的内容（Delta元素）

1.4 Discussion Prompt

1.4 讨论问题

Ask participants: "What is the worst AI output you have seen or produced? What do you think was missing from the prompt?"
向参与者提问：“你见过或生成过的最差AI输出是什么？你认为提示词中缺少了什么？”

Module 2: The Alpha-Beta-Gamma-Delta-Epsilon Framework (45 minutes)

模块2：Alpha-Beta-Gamma-Delta-Epsilon框架（45分钟）

Based on Upadhyay (2024), Generative AI for Marketing, Chapter 3.
基于Upadhyay (2024)，《Generative AI for Marketing》，第3章。

Learning Objective

学习目标

Teach participants to construct a complete, high-quality marketing prompt using the five-element Alpha-Beta-Gamma-Delta-Epsilon structure.
教授参与者使用Alpha-Beta-Gamma-Delta-Epsilon五元素结构构建完整、高质量的营销提示词。

2.1 The Five Elements

2.1 五大元素

Every high-quality marketing prompt contains five components. Walk through each one.
Alpha — Role Assignment
Assign an expert persona to the AI. The role anchors the AI's output register, expertise level, and perspective.
Formula: "You are a [role] who specialises in [specialism] for [market/industry]."
Example: "You are a senior copywriter who specialises in social media content for Ugandan financial services brands."
Why it matters: Without a role, the AI defaults to a generic, cautious register. With a strong role, output is confident, specific, and appropriate to the industry.
Beta — Context Block
Provide all relevant background. Include brand voice, target audience, platform, campaign background, and any specific product or offer details.
Formula: Brand name + audience demographics + platform + tone/voice + product or offer details.
Example: "The brand is [X], targeting urban professionals aged 25–35 in Kampala. Brand voice: confident, friendly, slightly witty. Product: 5G data bundle, UGX 15,000 for 5GB/7 days."
Why it matters: Thin context produces generic output. The more specific the Beta, the more tailored the result.
Gamma — Task Instruction
State exactly what to produce. Include format, length, quantity, perspective, and any framework to apply.
Formula: "Write [quantity] [format] using [framework]. Each must be [length constraint] and include [required elements]."
Example: "Write 3 Instagram captions using the PAS framework (Problem → Agitate → Solution). Each caption must be under 150 characters, end with a clear CTA, and include 2–3 relevant hashtags."
Why it matters: Without a clear task instruction, the AI interprets ambiguity in its own favour — often producing the easiest, most generic interpretation.
Delta — Constraints
State explicitly what to avoid: banned words, competitor mentions, topics, tone characteristics, or content types.
Formula: "Do not [X]. Do not use the word/phrase [Y]. Avoid [Z]."
Example: "Do not mention competitor brands. Do not use the words 'affordable' or 'seamless'. Avoid exclamation marks."
Why it matters: AI tools default to common marketing clichés. Constraints remove the most predictable and weakest outputs.
Epsilon — Output Format
Specify exactly how to present the result: numbered list, table, plain paragraph, JSON, labelled sections, etc.
Formula: "Present as [format]. Label each [element] with [label]."
Example: "Present as a numbered list with the PAS label for each element."
Why it matters: Formatting instructions make output paste-ready and prevent the AI from burying the useful content inside paragraphs of explanation.
每一条高质量的营销提示词都包含五个组件。逐一讲解。
Alpha——角色分配
为AI分配专家角色。该角色决定AI输出的语体、专业水平和视角。
公式： “你是一名[角色]，专门从事[领域]的[市场/行业]工作。”
示例： “你是一名资深文案师，专门为乌干达金融服务品牌创作社交媒体内容。”
重要性： 若未指定角色，AI会默认使用通用、谨慎的语体。明确清晰的角色后，输出内容会更自信、具体，且符合行业需求。
Beta——背景模块
提供所有相关背景信息。包括品牌调性、目标受众、平台、活动背景，以及任何特定产品或优惠详情。
公式： 品牌名称 + 受众人口统计信息 + 平台 + 语气/调性 + 产品或优惠详情。
示例： “品牌为[X]，目标受众是坎帕拉25-35岁的城市专业人士。品牌调性：自信、友好、略带诙谐。产品：5G流量包，15000乌干达先令，5GB/7天。”
重要性： 背景信息不足会导致输出内容通用化。Beta元素越具体，结果就越贴合需求。
Gamma——任务指令
明确说明需要生成的内容。包括格式、篇幅、数量、视角，以及需应用的框架。
公式： “撰写[数量]份[格式]内容，使用[框架]。每份内容需满足[篇幅限制]，并包含[必填元素]。”
示例： “撰写3条使用PAS框架（问题→激化→解决方案）的Instagram标题文案。每条文案不超过150字符，结尾包含清晰的行动号召，并添加2-3个相关话题标签。”
重要性： 若任务指令不清晰，AI会自行解读歧义——通常生成最简单、最通用的内容。
Delta——限制条件
明确说明需规避的内容：禁用词汇、竞品提及、话题、语气特征或内容类型。
公式： “请勿[X]。请勿使用词汇/短语[Y]。避免[Z]。”
示例： “请勿提及竞品品牌。请勿使用‘实惠’或‘流畅’等词汇。避免使用感叹号。”
重要性： AI工具默认会使用常见的营销陈词滥调。限制条件可剔除最具预测性、质量最差的输出内容。
Epsilon——输出格式
明确说明结果的呈现方式：编号列表、表格、普通段落、JSON、带标签的章节等。
公式： “以[格式]呈现。为每个[元素]添加[标签]。”
示例： “以编号列表呈现，为每个元素添加PAS标签。”
重要性： 格式说明可让输出内容直接复制使用，避免AI将有用内容隐藏在解释性段落中。

2.2 The 10 Prompt Components

2.2 10个提示词组件

Expand the framework into the full 10 components that can appear in a complete marketing prompt (Upadhyay, 2024):
#ComponentDescription
1Role assignmentExpert persona for the AI (Alpha)
2Audience descriptionDemographics, psychographics, location
3Platform / channelInstagram, WhatsApp, email, LinkedIn, etc.
4ObjectiveAwareness / engagement / conversion / retention
5Tone instructionFormal, friendly, witty, urgent, warm, etc.
6Length / formatCharacter count, word count, list vs. paragraph
7ConstraintsWhat to avoid — words, topics, brands
8Context blockBrand, product, campaign background (Beta)
9Few-shot examples2–3 examples of the desired output style
10Output structureHow to present the final result (Epsilon)
Explain that not every prompt requires all 10 components. Simple tasks may need only 4–5. Complex, high-stakes content (campaign launches, CEO communications) should use all 10.
将框架扩展为完整的10个组件，可用于构建完整的营销提示词（Upadhyay, 2024）：
#组件描述
1角色分配AI的专家角色（Alpha）
2受众描述人口统计、心理统计、地理位置
3平台/渠道Instagram、WhatsApp、邮件、LinkedIn等
4目标认知/互动/转化/留存
5语气指令正式、友好、诙谐、紧迫、温暖等
6篇幅/格式字符数、字数、列表或段落
7限制条件需规避的内容——词汇、话题、品牌
8背景模块品牌、产品、活动背景（Beta）
9少量示例2-3个理想输出风格的示例
10输出结构最终结果的呈现方式（Epsilon）
说明并非所有提示词都需要包含全部10个组件。简单任务可能只需4-5个。复杂、高风险的内容（活动启动、CEO沟通）应包含全部10个组件。

2.3 Common Mistakes

2.3 常见错误

Walk through the four most frequent prompt errors:
  1. Role missing — output lacks expertise and authority; reads like generic text
  2. Context too thin — output is generic, uses placeholder language, misses brand voice
  3. No constraints — output is full of clichés: "seamless", "affordable", "game-changing"
  4. No format instruction — output is buried in paragraphs; hard to use without reformatting
讲解四种最常见的提示词错误：
  1. 缺少角色——输出内容缺乏专业性和权威性；读起来像通用文本
  2. 背景信息不足——输出内容通用化，使用占位符语言，不符合品牌调性
  3. 无限制条件——输出内容充满陈词滥调：“流畅”“实惠”“颠覆性”
  4. 无格式说明——输出内容隐藏在段落中；需重新格式化才能使用

2.4 Hands-On Activity (15 minutes)

2.4 实操活动（15分钟）

Ask participants to build a complete Alpha-Beta-Gamma-Delta-Epsilon prompt for their own brand. Provide a blank template:
Alpha (Role): You are a ___________________________________
Beta (Context): The brand is _____, targeting _____ aged _____ in _____. Brand voice: _____. Product/offer: _____.
Gamma (Task): Write _____ using the _____ framework. Each must be _____ and include _____.
Delta (Constraints): Do not _____. Do not use the word "_____". Avoid _____.
Epsilon (Output format): Present as _____.
Participants submit their prompt to the AI tool and share their output with the group.
让参与者为自己的品牌构建完整的Alpha-Beta-Gamma-Delta-Epsilon提示词。提供空白模板：
Alpha (角色): 你是一名 ___________________________________
Beta (背景): 品牌为_____，目标受众是_____，年龄_____，位于_____。品牌调性：_____。产品/优惠：_____。
Gamma (任务): 撰写_____，使用_____框架。每份内容需满足_____，并包含_____。
Delta (限制条件): 请勿_____。请勿使用词汇“_____”。避免_____。
Epsilon (输出格式): 以_____呈现。
参与者将提示词输入AI工具，并与小组分享输出内容。

Module 3: Copywriting Frameworks in Prompts (45 minutes)

模块3：提示词中的文案框架（45分钟）

Learning Objective

学习目标

Teach participants to embed copywriting frameworks in the Gamma element of a prompt, producing structured, persuasive marketing copy with consistent logic.
教授参与者在提示词的Gamma元素中嵌入文案框架，生成结构清晰、具有说服力的营销文案，逻辑一致。

3.1 Why Frameworks Matter in Prompts

3.1 框架在提示词中的重要性

Explain that copywriting frameworks give the AI a logical structure to follow. Without a framework instruction, AI copy tends to be flat and unconvincing. With a framework, output follows a proven persuasion sequence.
解释文案框架为AI提供了可遵循的逻辑结构。若未指定框架指令，AI生成的文案往往平淡无奇、缺乏说服力。使用框架后，输出内容会遵循经过验证的说服逻辑。

3.2 The Seven Frameworks

3.2 七大框架

Explain each framework with a Uganda/East Africa brand example. Show participants how to embed the framework name in the Gamma element.
1. PAS — Problem / Agitate / Solution
Structure: Identify the pain point → make it emotionally vivid → present the answer.
Prompt instruction (Gamma): "Write using the PAS framework: first name the problem the audience faces, then agitate it (make it feel urgent or frustrating), then present the product as the solution."
EA example: For a Kampala fuel delivery brand — Problem: running out of petrol at the wrong time. Agitate: stuck in Ntinda traffic with an empty tank, no fuel station in sight. Solution: [Brand] delivers fuel to your location in under 30 minutes.
2. AIDA — Attention / Interest / Desire / Action
Structure: Grab attention → build interest → create desire → prompt action.
Prompt instruction (Gamma): "Write using the AIDA framework: open with a bold attention hook, develop interest with relevant detail, build desire with a benefit or outcome, close with a direct call to action."
EA example: For a Ugandan savings app — Attention: "Every week you delay saving costs you more than you think." Interest: [App] automatically saves UGX 5,000 from every mobile money transaction. Desire: Thousands of Ugandans have saved their first UGX 1,000,000 without thinking about it. Action: Download and start saving today.
3. BAB — Before / After / Bridge
Structure: Show the painful before state → paint the desirable after state → bridge with the product.
Prompt instruction (Gamma): "Write using the BAB framework: describe the before state (audience's current frustration), describe the after state (what life looks like with the solution), then bridge the two with the product."
EA example: For a Nairobi laundry service — Before: spending your Sunday washing instead of resting. After: clean, folded, delivered by Saturday evening. Bridge: [Brand] handles your laundry so your Sunday is yours again.
4. FAB — Features / Advantages / Benefits
Structure: State the product feature → explain its advantage → describe the real benefit to the customer.
Prompt instruction (Gamma): "Write using the FAB framework: name the feature, state the functional advantage it provides, then articulate the emotional or practical benefit the customer experiences."
EA example: For a Ugandan solar home system — Feature: 200W solar panel with 10-hour battery backup. Advantage: powers lights, phone charging, and a small TV without grid electricity. Benefit: your household stays connected and your children can study after dark.
5. SSS — Star / Story / Solution
Structure: Introduce a relatable character (Star) → tell their struggle (Story) → resolve it with the product (Solution).
Prompt instruction (Gamma): "Write using the SSS framework: introduce a relatable character facing a problem, tell a short story of their struggle, then resolve it with the product or service."
EA example: For a Tanzanian health insurance brand — Star: Mama Zawadi, a market vendor in Dar es Salaam. Story: when her son fell ill, she had to close her stall for three days and borrow money she couldn't afford to repay. Solution: [Brand] covered the hospital bill. She never closed her stall again.
6. PPPP — Picture / Promise / Prove / Push
Structure: Create a vivid mental image → make a clear promise → provide evidence → push to action.
Prompt instruction (Gamma): "Write using the PPPP framework: paint a vivid picture of the desired outcome, make a specific promise, provide proof or social evidence, then push the reader to act now."
EA example: For a Ugandan real estate brand — Picture: imagine driving through your gate, not your landlord's. Promise: we help first-time buyers own a home in Wakiso from UGX 120M. Prove: over 400 families have moved in since 2021. Push: book your site visit this weekend — plots are releasing fast.
7. AFOREST — Alliteration / Facts / Opinions / Repetition / Examples / Rhetorical questions / Statistics / Three-part lists
Structure: A toolkit of persuasive writing devices — use selectively rather than all at once.
Prompt instruction (Gamma): "Write using AFOREST persuasive devices. Include at least: one factual statement, one rhetorical question, one statistic, and a three-part list."
EA example instruction to AI: "You are a copywriter for a Kenyan mobile data brand. Write a Facebook post using AFOREST devices: include a rhetorical question, a statistic about mobile data usage in Kenya, and a three-part list of what customers can do with the data bundle."
逐一讲解每个框架，并搭配乌干达/东非品牌示例。向参与者展示如何在Gamma元素中嵌入框架名称。
1. PAS——问题/激化/解决方案
结构： 识别痛点→强化情绪共鸣→给出解决方案。
提示词指令（Gamma）： “使用PAS框架撰写：首先点明受众面临的问题，然后激化问题（使其显得紧迫或令人沮丧），最后将产品作为解决方案呈现。”
东非示例： 针对坎帕拉燃油配送品牌——问题：在错误的时间耗尽燃油。激化：被困在Ntinda的车流中，油箱空空，附近没有加油站。解决方案：[品牌]可在30分钟内将燃油送到你的位置。
2. AIDA——注意/兴趣/欲望/行动
结构： 吸引注意力→培养兴趣→激发欲望→促使行动。
提示词指令（Gamma）： “使用AIDA框架撰写：以醒目的注意力钩子开头，用相关细节培养兴趣，通过利益或成果激发欲望，以直接的行动号召结尾。”
东非示例： 针对乌干达储蓄应用——注意：“每周推迟储蓄都会让你付出超出想象的代价。”兴趣：[应用]会自动从每笔移动支付交易中扣除5000乌干达先令进行储蓄。欲望：数千乌干达人在无需刻意关注的情况下存下了人生中的第一个100万乌干达先令。行动：立即下载并开始储蓄。
3. BAB——之前/之后/桥梁
结构： 展示痛苦的现状→描绘理想的未来→用产品连接两者。
提示词指令（Gamma）： “使用BAB框架撰写：描述现状（受众当前的困扰），描述未来（使用解决方案后的生活状态），然后用产品连接两者。”
东非示例： 针对内罗毕洗衣服务——之前：周日花时间洗衣服，无法休息。之后：周六晚上就能收到干净、叠好的衣物。桥梁：[品牌]帮你处理洗衣事务，让周日完全属于你。
4. FAB——特性/优势/利益
结构： 说明产品特性→解释其优势→描述客户获得的实际利益。
提示词指令（Gamma）： “使用FAB框架撰写：点明特性，说明其带来的功能优势，然后阐述客户获得的情感或实际利益。”
东非示例： 针对乌干达家用太阳能系统——特性：200W太阳能电池板，10小时电池续航。优势：无需电网即可为灯光、手机充电和小型电视供电。利益：你的家庭保持连通，孩子可以在天黑后继续学习。
5. SSS——主角/故事/解决方案
结构： 引入一个相关的角色（主角）→讲述他们的困境（故事）→用产品解决问题（解决方案）。
提示词指令（Gamma）： “使用SSS框架撰写：引入一个面临问题的相关角色，讲述他们的简短困境故事，然后用产品或服务解决问题。”
东非示例： 针对坦桑尼亚健康保险品牌——主角：Mama Zawadi，达累斯萨拉姆的市场商贩。故事：她儿子生病时，她不得不关闭摊位三天，借了无力偿还的钱。解决方案：[品牌]支付了医药费。她再也不用关闭摊位了。
6. PPPP——描绘/承诺/证明/推动
结构： 创建生动的心理画面→做出明确承诺→提供证据→促使行动。
提示词指令（Gamma）： “使用PPPP框架撰写：描绘理想结果的生动画面，做出具体承诺，提供证据或社交证明，然后促使读者立即行动。”
东非示例： 针对乌干达房地产品牌——描绘：想象开车驶入自己的大门，而不是房东的。承诺：我们帮助首次购房者在Wakiso拥有一套售价1.2亿乌干达先令起的住房。证明：自2021年以来，已有400多个家庭入住。推动：本周末预约实地看房——地块正在快速推出。
7. AFOREST——头韵/事实/观点/重复/示例/反问/统计数据/三分列表
结构： 一套说服性写作工具——选择性使用，无需全部应用。
提示词指令（Gamma）： “使用AFOREST说服工具撰写。至少包含：一个事实陈述、一个反问句、一个统计数据和一个三分列表。”
东非示例（给AI的指令）： “你是一名肯尼亚移动数据品牌的文案师。使用AFOREST工具撰写一篇Facebook帖子：包含一个反问句、一个关于肯尼亚移动数据使用情况的统计数据，以及一个关于客户可使用流量包做什么的三分列表。”

3.3 Embedding Frameworks in the Gamma Element

3.3 在Gamma元素中嵌入框架

Show participants the exact pattern for embedding any framework in a prompt:
"Write [quantity] [content type] using the [FRAMEWORK NAME] framework ([brief description of the framework sequence]). [Add length, CTA, and hashtag requirements here.]"
向参与者展示在提示词中嵌入任何框架的具体模式：
“撰写[数量]份[内容类型]，使用[框架名称]框架（[框架逻辑的简要描述]）。[在此添加篇幅、行动号召和话题标签要求。]”

3.4 Hands-On Activity (20 minutes)

3.4 实操活动（20分钟）

Participants complete two exercises:
Exercise A: Write one Instagram caption using the PAS framework for their own brand. Run it through the AI, review the output, and refine once.
Exercise B: Write one email subject line using the AIDA framework for their own brand. Run it through the AI, compare 3 variations.
Share and discuss outputs with the group.
参与者完成两个练习：
练习A： 为自己的品牌撰写一条使用PAS框架的Instagram标题文案。将其输入AI，查看输出内容并优化一次。
练习B： 为自己的品牌撰写一条使用AIDA框架的邮件主题行。将其输入AI，比较3个版本。
与小组分享并讨论输出内容。

Module 4: Practical Prompt Library and Iterative Refinement (30 minutes)

模块4：实用提示词库与迭代优化（30分钟）

Learning Objective

学习目标

Equip participants with a reusable set of prompt templates and the habit of iterative refinement — so they leave training with tools they can use the next day.
为参与者提供可复用的提示词模板集，并培养迭代优化的习惯——让他们在培训结束后就能使用这些工具开展工作。

4.1 The Prompt Library

4.1 提示词库

Walk through 7 ready-made prompt templates from the 
prompt-engineering-library
skill. Cover one template from each common content type:
  1. Instagram / Facebook caption
  2. WhatsApp broadcast message
  3. Email subject line
  4. Email body (short-form)
  5. LinkedIn post (thought leadership)
  6. Blog post introduction
  7. SMS / short copy
For each template, show participants where to substitute their own brand details. Demonstrate live on the AI tool.
Reference: For the full library of templates, see the 
prompt-engineering-library
skill.
讲解
prompt-engineering-library
技能中的7个现成提示词模板。涵盖每种常见内容类型的一个模板：
  1. Instagram/Facebook标题文案
  2. WhatsApp广播消息
  3. 邮件主题行
  4. 邮件正文（短格式）
  5. LinkedIn帖子（思想领导力）
  6. 博客文章引言
  7. 短信/短文案
针对每个模板，向参与者展示如何替换自己的品牌信息。在AI工具上进行现场演示。
参考： 完整模板库，请查看
prompt-engineering-library
技能。

4.2 Iterative Refinement

4.2 迭代优化

Explain the five prompting approaches (Upadhyay, 2024) and when to use each:
ApproachDescriptionBest for
Zero-shotSingle instruction, no examplesSimple tasks: date, format, quick edits
One-shotOne example provided before the requestTone matching when you have one good example
Few-shot2–5 examples provided before the requestBrand voice matching; consistent series of posts
Chain-of-thoughtAsk AI to reason step by step before outputtingStrategy tasks: campaign planning, audience analysis
Iterative refinementStart broad, narrow with follow-up promptsAny high-stakes content requiring multiple drafts
Demonstrate iterative refinement live using a three-prompt sequence:
  • Prompt 1 (broad): Run the initial Alpha-Beta-Gamma-Delta-Epsilon prompt. Review output.
  • Prompt 2 (narrow): "Good. Now shorten the first caption to under 100 characters and make the CTA more direct. Keep everything else the same."
  • Prompt 3 (refine): "Replace the third hashtag with a more location-specific Ugandan tag. Do not change the rest."
Show that iterative refinement produces better results than trying to write the perfect prompt first time.
讲解五种提示方法（Upadhyay, 2024）及适用场景：
方法描述最佳适用场景
零样本（Zero-shot）单一指令，无示例简单任务：日期、格式、快速编辑
单样本（One-shot）请求前提供一个示例当你有一个优质示例时，用于匹配语气
少样本（Few-shot）请求前提供2-5个示例匹配品牌调性；生成一致的系列帖子
思维链（Chain-of-thought）要求AI先逐步推理，再输出内容策略任务：活动规划、受众分析
迭代优化（Iterative refinement）从宽泛指令开始，通过后续提示缩小范围任何需要多轮草稿的高风险内容
现场演示迭代优化，使用三步提示序列：
  • 提示1（宽泛）： 运行初始的Alpha-Beta-Gamma-Delta-Epsilon提示词。查看输出内容。
  • 提示2（缩小范围）： “很好。现在将第一条标题文案缩短至100字符以内，让行动号召更直接。其他内容保持不变。”
  • 提示3（优化）： “将第三个话题标签替换为更贴合乌干达本地的标签。不要更改其他内容。”
展示迭代优化比一次性写出完美提示词能产生更好的结果。

4.3 The AI Content Quality Checklist

4.3 AI内容质量检查表

Before any AI-generated content is published, apply the checklist from the 
ai-content-humaniser
skill. Key checks include:
  • Does this sound like a human wrote it, or an AI?
  • Does it match the brand voice guide?
  • Are all facts, statistics, and figures verified?
  • Are there any AI clichés ("game-changing", "seamless", "dive into", "in today's fast-paced world")?
  • Has a human editor reviewed and approved it?
Reference: For the full humanisation checklist and editing protocol, see the 
ai-content-humaniser
skill.
在发布任何AI生成的内容前，应用
ai-content-humaniser
技能中的检查表。关键检查项包括：
  • 读起来像人类写的，还是AI生成的？
  • 是否符合品牌调性指南？
  • 所有事实、统计数据和数字是否经过验证？
  • 是否存在AI陈词滥调（“颠覆性”“流畅”“深入探索”“在当今快节奏的世界中”）？
  • 是否经过人工编辑审核和批准？
参考： 完整的人性化检查表和编辑流程，请查看
ai-content-humaniser
技能。

4.4 East African Context Notes

4.4 东非注意事项

Apply these additional considerations when prompting for EA-market content:
  • WhatsApp broadcast prompts require different constraints than email or Instagram. Specify: conversational tone, short paragraphs, no HTML, no markdown formatting, clear opt-out line.
  • Luganda / Swahili phrases can be embedded. Instruct the AI: "Include one Luganda phrase in the caption and provide an English translation in brackets."
  • Low-data context means shorter, punchier copy performs better. Default to under 150 characters for captions; under 160 characters for SMS.
  • Local cultural context is powerful. Instruct the AI to reference locally recognisable elements: boda-boda, rolex (the Ugandan street food), market day, supper culture, Kampala traffic, KCCA, etc.
  • UGX pricing should always appear in prompts where price is relevant. Do not let the AI invent pricing — always provide it in the Beta element.
为东非市场内容撰写提示词时，需考虑以下额外因素：
  • WhatsApp广播提示词需要与邮件或Instagram不同的限制条件。明确说明：口语化语气、短段落、无HTML、无Markdown格式、清晰的退订说明。
  • 可嵌入卢干达语/斯瓦希里语短语。指示AI：“在标题文案中加入一个卢干达语短语，并在括号中提供英文翻译。”
  • 低流量环境意味着更简短、有力的文案效果更好。标题文案默认不超过150字符；短信不超过160字符。
  • 本地文化背景非常有效。指示AI参考本地可识别的元素：摩托车出租车（boda-boda）、rolex（乌干达街头美食）、集市日、晚餐文化、坎帕拉交通、KCCA等。
  • 若涉及价格，提示词中应始终使用乌干达先令（UGX）。请勿让AI编造价格——务必在Beta元素中提供。

4.5 Prompt Type Taxonomy and Examples-First Methodology

4.5 提示词类型分类与示例优先方法

Eight prompt types — use as a decision tool:
TypeWhen to Use
Zero-shotSimple tasks, quick edits
Few-shotBrand voice matching — correct default for any voice task
InstructionalComplex multi-part deliverables
ConversationalRefinement, ideation, brainstorming
ContextualStrategy documents, personas, reports
CreativeStorytelling, campaign concepts
Chain-of-ThoughtStrategy, audience analysis, decision-making
SystematicRepeatable tasks: caption series, email sequences
Few-shot is the correct default for brand voice work. Supply 2–3 examples of actual human writing before issuing a content task. Master structure:
Here is an example of [content type] in the voice and style I want:
###
[Example 1]
###
[Example 2]
###
Now write a new [content type] on [topic] in the same voice and style.
Adjectives are imprecise; examples are exact. Three or more examples enable reliable voice matching; fewer produce inconsistency. Training activity: participants paste three real brand posts, generate a new post, compare against tone-adjective-only output.
八种提示词类型——用作决策工具：
类型适用场景
零样本简单任务、快速编辑
少样本匹配品牌调性——任何调性任务的正确默认选择
指令型复杂的多部分交付成果
对话型优化、构思、头脑风暴
情境型策略文档、用户画像、报告
创意型故事讲述、活动概念
思维链策略、受众分析、决策制定
系统型可重复任务：标题文案系列、邮件序列
少样本是品牌调性工作的正确默认选择。在下达内容任务前，提供2-3个真实的人工撰写示例。掌握以下结构：
以下是我想要的[内容类型]的语气和风格示例：
###
[示例1]
###
[示例2]
###
现在以相同的语气和风格撰写一篇关于[主题]的新[内容类型]。
形容词不够精确；示例才是准确的。3个及以上示例可实现可靠的调性匹配；示例越少，结果越不一致。培训活动：参与者粘贴3条真实品牌帖子，生成新帖子，与仅使用形容词描述调性的输出进行比较。

Worked Prompt Examples

实操提示词示例

Example 1: Instagram Caption — PAS Framework (Telecoms Brand)

示例1：Instagram标题文案——PAS框架（电信品牌）

Alpha: You are a senior social media copywriter specialising in Ugandan telecoms brands.

Beta: The brand is [X], targeting urban professionals aged 25–35 in Kampala. Brand voice: confident, friendly, slightly witty. Product: 5G data bundle, UGX 15,000 for 5GB/7 days.

Gamma: Write 3 Instagram captions using the PAS framework (Problem → Agitate → Solution). Each caption must be under 150 characters, end with a clear CTA, and include 2–3 relevant hashtags.

Delta: Do not mention competitor brands. Do not use the words "affordable" or "seamless". Avoid exclamation marks.

Epsilon: Present as a numbered list with the PAS label for each element.
Alpha: 你是一名资深社交媒体文案师，专门为乌干达电信品牌创作内容。

Beta: 品牌为[X]，目标受众是坎帕拉25-35岁的城市专业人士。品牌调性：自信、友好、略带诙谐。产品：5G流量包，15000乌干达先令，5GB/7天。

Gamma: 撰写3条使用PAS框架（问题→激化→解决方案）的Instagram标题文案。每条文案不超过150字符，结尾包含清晰的行动号召，并添加2-3个相关话题标签。

Delta: 请勿提及竞品品牌。请勿使用“实惠”或“流畅”等词汇。避免使用感叹号。

Epsilon: 以编号列表呈现，为每个元素添加PAS标签。

Example 2: Email Subject Lines — AIDA Framework (Microfinance Institution)

示例2：邮件主题行——AIDA框架（小额信贷机构）

Alpha: You are an email copywriter for an East African microfinance institution.

Beta: We are sending a re-engagement email to customers who have not transacted in 90 days. Brand tone: warm, encouraging, professional.

Gamma: Write 5 subject line options using the AIDA framework (Attention → Interest → Desire → Action), each under 50 characters.

Delta: Do not mention overdue balances. Do not use the word "urgent".

Epsilon: Numbered list, subject line only, no explanation.
Alpha: 你是一名为东非小额信贷机构服务的邮件文案师。

Beta: 我们将向90天未交易的客户发送一封重新激活邮件。品牌语气：温暖、鼓励、专业。

Gamma: 撰写5个使用AIDA框架（注意→兴趣→欲望→行动）的主题行选项，每个不超过50字符。

Delta: 请勿提及逾期余额。请勿使用“紧急”一词。

Epsilon: 编号列表，仅包含主题行，无解释。

Example 3: WhatsApp Broadcast — BAB Framework (Delivery Service)

示例3：WhatsApp广播——BAB框架（配送服务）

Alpha: You are a WhatsApp copywriter for a Kampala-based same-day delivery service.

Beta: Brand: [X]. Audience: small business owners and traders aged 28–45 in Kampala. Tone: direct, reliable, no-nonsense. Promotion: free delivery on orders above UGX 50,000 this Friday only.

Gamma: Write one WhatsApp broadcast message using the BAB framework (Before → After → Bridge). Maximum 3 short paragraphs. End with a WhatsApp reply CTA ("Reply YES to book your first delivery").

Delta: No emojis. No markdown. No bullet points — plain text only. Do not use the phrase "limited time".

Epsilon: Single plain-text message, no formatting, ready to paste into WhatsApp Business.
Alpha: 你是一名为坎帕拉当日配送服务服务的WhatsApp文案师。

Beta: 品牌：[X]。受众：坎帕拉28-45岁的小企业主和商贩。语气：直接、可靠、务实。促销活动：本周五仅对5万乌干达先令以上的订单提供免费配送。

Gamma: 使用BAB框架（之前→之后→桥梁）撰写一条WhatsApp广播消息。最多3个短段落。结尾包含WhatsApp回复行动号召（“回复YES预订你的首次配送”）。

Delta: 无表情符号。无Markdown。无项目符号——仅纯文本。请勿使用“限时”一词。

Epsilon: 单一纯文本消息，无格式，可直接粘贴到WhatsApp Business中。

Quality Criteria

质量标准

The completed training guide meets the standard if:
  • All 4 modules are included with accurate time allocations totalling approximately 2.5 hours
  • The Alpha-Beta-Gamma-Delta-Epsilon framework is explained in full with at least one complete worked example per element
  • All 7 copywriting frameworks are named, defined, and illustrated with a Uganda/East Africa brand example
  • Hands-on activities are specified in Modules 2, 3, and 4 — not lecture content only
  • All worked examples use Ugandan/EA brands, UGX pricing, and local cultural references where relevant
  • Output is structured so a non-technical facilitator can deliver it without additional preparation
  • The 
    prompt-engineering-library
    and 
    ai-content-humaniser
    skills are explicitly referenced as companion resources
  • The guide is written in British English with imperative language throughout
完成的培训指南符合以下标准：
  • 包含全部4个模块，时间分配准确，总时长约2.5小时
  • 完整讲解Alpha-Beta-Gamma-Delta-Epsilon框架，每个元素至少包含一个完整的实操示例
  • 命名、定义所有7个文案框架，并搭配乌干达/东非品牌示例
  • 模块2、3、4中明确指定实操活动——而非仅包含讲座内容
  • 所有实操示例使用乌干达/东非品牌、乌干达先令定价，并在相关情况下参考本地文化元素
  • 输出内容结构清晰，非技术主持人无需额外准备即可开展培训
  • 明确引用
    prompt-engineering-library
    和
    ai-content-humaniser
    技能作为配套资源
  • 使用英式英语撰写，全程使用祈使语气