Idea-to-Post Expansion Skill

Skill Overview

This skill helps you expand scattered ideas (a sentence, a few words, a vague thought) into 90+ point social media posts.

Target Positioning: Quality social media content, not technical documentation

It works through the following ways:

1. 1. Framework Internalization - Use thinking frameworks to design questions, but don't mechanically apply them
2. Information Search & Integration - Automatically search for high-quality materials, supplement relevant data and cases
3. Progressive Deep Questioning - Multiple-choice to set direction + open questions to enrich content, 7-10 rounds of dialogue until complete
4. Iterative Polishing & Optimization - Reflect and optimize after generation, pursuing 90+ point quality
5. Multi-Platform Output - Generate post versions adapted to different platforms

Expected dialogue rounds: 7-10 rounds

• - 3-4 rounds: Get direction and core viewpoints (technical documentation level)
• 5-7 rounds: Add cases, emotions, uniqueness (social media level)
• 7-10 rounds: Deep mining, repeated polishing (quality social media level)

Core Mechanism: Progressive Questioning + Framework Internalization

Three Core Principles

1. Internalize Frameworks, Don't Expose Them

Use thinking framework logic to design questions, but don't say "I'm using [Framework Name]":

CODEBLOCK0

2. Combine Multiple Choice + Open Questions

CODEBLOCK1

3. Progressive Deepening, Dynamic Adjustment

Each round of questions is based on the previous answer, naturally transitioning to the next dimension:

CODEBLOCK2

---

Questioning Flow (7-10 Rounds of Dialogue + Multi-Stage Search)

CODEBLOCK3

Note: Cases, emotions, and uniqueness are required and cannot be skipped. The second round of precision search is a key环节.

---

Integrating Framework Thinking into Questions

Framework Thinking	Questioning Approach	Example
Point	What's the core viewpoint?	"What's the core viewpoint you want to express?"
Reason

Why do you think so? | "Why do you think so? What's the reason?" |
| What | What specifically? | "What does this specifically refer to?" |
| Why | Why is it important? | "What pain point does it solve?" |
| How | How is it done? | "How is it implemented?" |
| Example | Any examples? | "Any specific cases?" |
| Situation | Initial state? | "What was the initial state?" |
| Complication | What conflict? | "What challenge appeared?" |

Frameworks are thinking tools, not questioning templates.

---

Quick Framework Selection Guide

Automatic Recommendation Rules

Based on keywords in your input, the system will automatically recommend frameworks:

Keywords	Recommended Framework
Why, essence, original intention, mission, value	Golden Circle
Problem, challenge, dilemma, turning point, story

SCQA |
| Promotion, publicity, conversion, sales, marketing | AIDA |
| Viewpoint, opinion, think, should, suggest | PREP |
| Deep dive, root cause, trace back, underlying | 5-Why |
| Innovation, breakthrough, disrupt, reconstruct, essence | First Principles |
| Product, feature, advantage, selling point, characteristic | FBA |
| Other or unclear | 5W1H (default) |

Framework Introduction

For detailed framework explanations, refer to references/thinking-frameworks.md

---

Usage Flow

Core Flow: Keep Questioning Until Complete

CODEBLOCK4

Information Search (Multi-Stage Execution, Required)

Search must be performed in multiple stages to ensure precise external validation is collected:

Stage 1: Initial Background Search (Before questioning begins)

Execute initial information search before questioning starts:

1. 1. Identify core concepts in the topic
2. Build multi-angle search queries
3. Use WebSearch to get relevant materials
4. Use WebFetch to read key pages in depth
5. Organize search results for later use

Stage 2: Precision Deep Search (After collecting core viewpoints)

This is the most critical step - after collecting the user's core viewpoints and unique insights, you must perform a second round of precision search:

1. 1. Build search terms based on user's unique viewpoints

- Extract unique insights/counter-intuitive viewpoints proposed by user - Use these viewpoints as keywords for reverse search - Look for supporting or refuting evidence

1. 2. Multi-angle validation search

- Search industry reports, news media - Search competitors/international cases - Search data support

1. 3. Handling search failures

- If one search fails/is limited, try different keyword combinations - Use more generic or more specific search terms - Try English search terms for international perspective - Record search status, inform user (e.g., if search was restricted)

Search Keyword Strategy

Stage	Search Focus	Example Keywords
Initial search	Background info, basic facts	"Qianwen 3 release", "AI e-commerce assistant"
Precision search

User unique viewpoint validation | "AI vs e-commerce conflict", "traffic distribution AI impact" | | Comparison search | International cases, competitor analysis | "ChatGPT e-commerce", "foreign AI shopping assistant" |

Search results must be integrated into final content as external validation.

Even if search is limited, try multiple different keywords and inform user of search status.

Question Design Principles

1. Question Based on Topic

Questions must be closely tied to the core topic of user input, don't ask irrelevant questions.

2. Framework-Based Design

Use framework thinking to design questions, but don't say "I'm using [Framework Name]".

3. Mix Multiple Choice + Open Questions

• - Multiple Choice: When there are clear options, need quick classification
• Open Questions: When need stories/experiences/emotions/details
• Mixed: After AskUserQuestion "Other" option, continue follow-up questions

4. Progressive Deepening, Dynamic Adjustment

Each round is based on the previous answer, naturally transitioning to the next dimension. Not mechanically following a template.

5. Focus on 1-2 Questions Per Round

Avoid information overload, give user thinking space.

6. Complete Information Before Stopping

Check core dimensions, question what's missing, only generate when complete.

Completeness Judgment Standards (90+ Point Target)

When judging whether information is complete, the skill checks the following dimensions:

Must Be Complete (Continue questioning if missing, cannot skip)

Dimension	Check Item	Description
Core Viewpoint	Is the core viewpoint to be expressed clear?	The soul of the article
Target Audience

Is it clear who it's written for? | Determines expression style | | Publishing Platform | Is it clear where to publish? | Determines content format | | Real Cases | Are there specific examples/experiences? | Social media required | | Emotional Resonance | Are there resonance points/emotional hooks? | Social media required | | Unique Viewpoints | Are there insights others haven't mentioned? | 90+ point required |

Should Be Complete (Try to question)

Dimension	Check Item	Description
External Validation	Is there search material to support?	Adds persuasiveness

Nice to Have (Better if present)

Dimension	Check Item	Description
Interactive Design	Is there a clear call to action?	Guide reader participation
Style Preference

What style? | Professional/humorous/story-based |

---

90+ Point Content Standards

Score	Characteristics	What's Missing
60-70 points	Clear structure, clear viewpoints	Lacks real cases, emotional resonance
80-85 points

+ Real cases, emotional resonance | Lacks uniqueness, external validation | | 90+ points | + Unique viewpoints, external validation | Nothing missing, polished |

Only enter generation phase when all "must complete" dimensions are present.

Real cases, emotional resonance, and unique viewpoints are the three pillars of social media content - all are essential.

---

Output Format

Markdown Format Specifications (Required)

All post content must be output in standard Markdown format, including:

Format Element	Use Case	Example
Heading Levels	Main title H1, sections H2-H4	INLINECODE1 INLINECODE2
Bold Emphasis

Core viewpoints, keywords | **Core viewpoint** |
| Quote Blocks | Golden sentences, key assertions | > Quote content |
| Lists | Parallel points, step descriptions | - Item 1 |
| Horizontal Rules | Separate different parts | --- |
| Code Blocks | Technical content, data | `code` |

Pre-output checklist:

• - [ ] Clear heading levels (H1 main title, H2 sections, H3 subsections)
• [ ] Bold emphasis on core viewpoints
• [ ] Quote blocks for golden sentences/key assertions
• [ ] Long content in bullet points
• [ ] Horizontal rules between sections
• [ ] Overall format follows standard Markdown syntax

Universal Structure

CODEBLOCK5

Platform-Adapted Versions

Platform	Word Count	Characteristics
WeChat Official Account	2000+	In-depth long articles, clear sections, image suggestions
Xiaohongshu

500-1000 | Practical content, emoji embellishment, list-style | | Twitter/Weibo | 140-280 | Concise and powerful, one-sentence core, golden sentence style | | LinkedIn/Maimai | 1000-1500 | Professional workplace, industry insights, case support |

For detailed structure explanations, refer to references/post-structures.md

---

High-Quality Information Sources

The system prioritizes the following types of high-quality sources when searching:

Source Type	Examples
Academic Resources	arXiv, Google Scholar, CNKI
Industry Reports

McKinsey, Gartner, iResearch |
| Professional Technical | Official docs, tech blogs, GitHub |
| News Media | Caixin, 36Kr, TechCrunch |
| Knowledge Platforms | Wikipedia, Zhihu high-voted, Medium |

For detailed data source lists, refer to references/data-sources.md

---

Best Practices

1. 1. Provide sufficient context - Even for scattered ideas, try to include key points you care about
2. Answer questions honestly - Your answers during interactive questioning directly affect post quality
3. Define target platform - Knowing which platform you're posting to can generate more suitable content
4. Leverage search results - Materials searched by the system can greatly enrich your content
5. Compare multiple versions - Compare outputs from different versions, choose the most suitable
6. Iterate and optimize - Continue questioning and optimizing based on generated results

---

Examples

See examples/ directory for complete usage examples:

• - basic-usage.md - Basic usage examples
• INLINECODE14 - Advanced scenario examples

---

Reference Documents

• - references/thinking-frameworks.md - Detailed framework explanations
• INLINECODE16 - Continuous progressive questioning strategy (core)
• INLINECODE17 - Questioning mode selection guide (new)
• INLINECODE18 - Question template library
• INLINECODE19 - Post structure guide
• INLINECODE20 - High-quality data source list