Purpose
Transform any expository/straightforward technical document into a hardcore article with hooks, rhythm, data, and no fluff.
Progressive Disclosure
- SKILL.md(this file) → General rewriting framework and workflow
- references/<style>/instructions.md → Operational instructions for a specific style (six-part arc, vocabulary conversion, self-check list, etc.)
- references/<style>/compose.md → Complete writing style breakdown (including example sentences, formulas)
- references/<style>/example.md → Reference original text, used for comparative learning
Available Styles
Currently supported styles:
| Style | Path | Core Features |
|---|---|---|
| Avi Chawla(Daily Dose of DS) | references/avi-chawla/ |
Counterintuitive hooks, data impact, single-sentence paragraphs, zero transitional phrases, reason reversal, contrast verification |
| Lionad Blog | references/lionad/ |
Colloquial troubleshooting narrative, heavy cuts leaving only the main thread, one technical mechanism explained thoroughly, problem-oriented subheadings driving progression |
If the user requests another style (such as Paul Graham, Nat Friedman, etc.), first analyze the reference article provided by the user, then under references/ create a corresponding directory and write instructions.md、compose.md and example.md。
Workflow
Step 1: Confirm the Writing Style
Before formally starting the rewrite, confirm the user's style direction:
If the user explicitly specified a style(such as "use Avi Chawla style"):
- Directly read
references/<style>/instructions.md, continue to Step 2
If the user did not specify a style:
- Show the list of available styles and their core features
- Let the user choose or provide a reference article URL
- If the user provides a specific reference article URL, prioritize analyzing that article, and under
references/create a new style directory
If the user only vaguely says "make it more impactful":
- Use Avi Chawla(Daily Dose of DS) style by default
- Show the user a style summary, and start only after confirmation
Step 2: Analyze the Original Text
Before rewriting, first perform the following analysis on the original text:
- Extract the core arguments: What are the 1-3 core points the original text tries to convey?
- Identify data/evidence: What numbers, cases, and comparisons are in the original text?
- Locate the problem/pain point: What problem does the original text describe? Why should readers care?
- Mark content that can be cut: Which transitional phrases, pleasantries, and repeated explanations can be deleted?
- Determine the language: Is the original text in Chinese or English? The target language should match the original text.
Step 3: Read and Apply the Style Guide
Read the corresponding style's instructions.md, and execute the rewrite according to the operational instructions in it.
If a deeper understanding of style details is needed, then read compose.md。
Step 4: Output Format
Directly output the full rewritten Markdown text.
Before outputting, first give a brief "rewriting strategy explanation" (2-3 sentences) telling the user the core argument you identified in the original text and how you plan to transform it.
Step 5: General self-check checklist
After completing the rewrite, perform the following general checks:
- Is the core argument preserved and made more prominent?
- Are the key data/evidence effectively used?
- The original text's image links
and links[text](url)Are all preserved? - Is the tone more direct and more impactful than the original text?
- Do all style-specific self-check checklists pass? (See
instructions.md)
Edge cases
- The original text is already in the target style: Prompt the user "The original text is already in the target style, no rewrite is needed," but you may provide fine-tuning suggestions
- The original text lacks data: If the original text truly has no quotable numbers, you may retain the descriptive argument, but try to transform abstract concepts into concrete scenarios
- The user requests another style: If the user explicitly requests another reference style, under
references/create a new directory, analyze the article provided by the user, and create the correspondinginstructions.md、compose.mdandexample.md - Non-technical documents: If it is a purely narrative/literary document, prompt the user that this style is mainly for technical writing and ask whether to continue