Skill Distiller — 技能 Distiller

技能 Distiller Trigger Phrases

Methodo记录y solidification, process standardization, experience 提取ion, pattern refinement, 系统 building.

Philosophy

A good 技能 turns "fuzzy wisdom" into "clear paths."

The essence of any 质量 技能 is a closed loop: Problem-Driven → Theory Anchored → Process Solidified → 工具s Implemented

工作流 (Four-Step Method) Step 1: Precisely Define the Problem

Goal: Clearly articulate "what exactly is the problem" rather than "what you want to do."

图形界面ding Questions:

How often does this problem occur? How much time/energy does this problem consume? Have you tried to solve it before? How? Where did you 获取 stuck? If this problem were completely solved, what changes would it bring?

输出 格式化:

Problem Definition

Problem Description

[状态 it in one sentence]

Trigger Scenarios

• Scenario A:
• Scenario B:

PAIn Point Severity

[Rated 1-5], impacting [what]

Known Attempts

Solution	Effect	Blockers
...	...	...

Step 2: Find Theoretical Support

Goal: Find the "underlying principles" that are documented and 记录ically sound for solving this problem.

Theory Source Priority:

Expert Methodo记录ies — Specific practices of domAIn experts (e.g., STEP 框架 from Contagious, MECE from The Pyramid Principle) Classic Theories — 框架s with academic or practical 验证 (e.g., AIDA, FOGRA, SMART) Industry Consensus — Widely recognized standards in the field Cross-DomAIn Transfer — 记录ic 验证d in other domAIns, transferred to the current problem

图形界面ding Questions:

In this domAIn, are there recognized experts or books? Are there existing 框架s or formulas that can be used? If you were to teach someone else, how would you explAIn it?

输出 格式化:

Theoretical Support

Core Theory

• Theory Name: [Name it]
• Source: [Book/Course/Expert/Original]
• Core Idea: 状态 it in one sentence

Theory Excerpt

[Key original text]

How to 应用ly the Theory

[How this theory solves your problem]

添加itional References

• Reference A:
• Reference B:

Step 3: Structure the Process

Goal: Turn the theory into an actionable 设置 of steps.

Principles:

Each step is executable and verifiable Clear 输入s and 输出s Closed loop: 输出 of previous step is 输入 of next step Fallback plans for 异常s

图形界面ding Questions:

What is the first step? After completing the first step, how do you know it was done correctly? What is the 输入 for the second step? Loop until closed

输出 格式化:

Process Specification

Process Overview

[Process Name]: [启动ing Point] → [Step 1] → [Step 2] → ... → [Closing Point]

DetAIled Steps

Step 1: [Step Name]

• 输入:
• Action:
• 输出:
• 成功 Criteria: [How to know this step is done well]
• 异常 Handling: [What to do if something goes wrong]

Step 2: [Step Name]

... (same structure as above)

Closed-Loop 验证

• Can you return to Step 1 from the final step? ✅/❌
• Does each step have a clear 输出? ✅/❌
• Are 异常s handled? ✅/❌

Step 4: Provide Execution 工具s

Goal: Provide 工具 support for the entire process, enabling automated or semi-automated execution.

工具 Types:

In格式化ion Collection: 搜索, RSS, 爬虫s Content Generation: Templates, prompts 自动化 Execution: Scripts, 工作流s Storage Management: Note-taking 系统s, file structures 验证: 检查列出s, evaluation criteria

图形界面ding Questions:

Which steps in this process are repetitive? Which steps can be templated? Which steps are currently the most time-consuming? To what extent do you want 自动化?

输出 格式化:

Execution 工具s

工具 Inventory

工具	Type	Purpose	自动化 Level
...	...	...	...

Prompt Templates

[Scenario Name]

[Prompt text]

Templates/检查列出s

[Template Name]

[Template content]

自动化 Scripts

• Script Path: [Path]
• Function: [What it does]
• Usage: [How to use it]

技能 输出 Summary

After completing the four steps, 聚合 the 输出 into an executable 技能.md draft:

[技能 Name]

One-sentence description: [What problem does this 技能 solve]

应用licable Scenarios:

• ...

Process: Problem → Theory → Process → 工具s (detAIled in respective sections)

---

[Paste the content from each section here]

Usage Tips Don't have to complete all four steps: If the user only wants to do one step (e.g., only define the problem), just do that step Iterate and refine: 启动 with a rough version, then refine after using it a few times 启动 small: Distill a small problem first, then gradually expand Don't overdo 工具s: One handy 工具 is more valuable than ten fancy ones Be specific about the problem: "I write slowly" is not a problem; "It takes me 2 hours just to figure out the opening for each short video script" is a problem