self-critique-writer

当用户请求“带着自我批评写作”、“批评并重写”、“改进这篇草稿”、“用我的风格迭代写作”,或希望内容生成循环通过草稿/批评/重写直至达到质量阈值时,应使用此技能。

📖 中文导读:当用户请求“带着自我批评写作”、“批评并重写”、“改进这篇草稿”、“用我的风格迭代写作”,或希望内容生成循环通过草稿/批评/重写直至达到质量阈值时,应使用此技能。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: self-critique-writer description: This skill should be used when the user asks to "write with self-critique", "critique and rewrite", "improve this draft", "write in my voice with iteration", or wants content generation that loops through draft/critique/rewrite until quality threshold is met. version: 1.0.0

Self-Critique Voice Writer

Generate content that matches your voice through iterative self-critique. Draft, score, critique, rewrite - repeat until the output is genuinely good.

The Problem with One-Shot Generation

Single-pass AI content is often:

  • Generic and forgettable
  • Missing your authentic voice
  • "Good enough" but not great
  • Obviously AI-generated

The Self-Critique Loop

[Voice Profile] → Draft → Score → Critique → Rewrite
                           ↑                    │
                           └────────────────────┘
                          (loop until score > 8)

How to Use

Basic Usage

/critique-write [content type] about [topic]

Example:

/critique-write linkedin post about the importance of code reviews

With Voice Reference

/critique-write twitter thread about AI agents, matching the voice from my recent posts

With Specific Critique Focus

/critique-write blog intro about remote work, focus critique on engagement and authenticity

The Critique Dimensions

Each draft is scored 1-10 on five dimensions:

1. Voice Match (Does it sound like me?)

  • Vocabulary alignment with profile
  • Sentence rhythm matching
  • Personality markers present
  • Avoids words on "never use" list

2. Clarity (Is it understandable?)

  • Main point obvious within 3 seconds
  • No jargon without context
  • Logical flow between ideas
  • Appropriate complexity for audience

3. Engagement (Will people read this?)

  • Hook strength (first line)
  • Tension or curiosity created
  • Momentum maintained
  • Satisfying resolution/CTA

4. Authenticity (Does it feel real?)

  • Specific over generic
  • Personal experience woven in
  • Honest about tradeoffs
  • Not "trying too hard"

5. Purpose Alignment (Does it achieve the goal?)

  • Clear intended outcome
  • Right audience targeted
  • Action prompted if applicable
  • No mixed messages

Example Session

User: /critique-write linkedin post about why I stopped using to-do lists

Draft 1:

I stopped using to-do lists 6 months ago. Here's what happened.

I used to be a to-do list fanatic. Multiple apps, color coding, priority levels. But I realized I was spending more time managing my lists than doing the work.

Now I use a simple rule: If it matters, I do it now. If it can wait, I write it on a sticky note. The sticky notes pile up. That's fine. Most of them become irrelevant before I get to them.

Sometimes the best productivity system is no system at all.

Self-Critique:

Dimension Score Notes
Voice Match 6 Too polished, missing your casual asides
Clarity 8 Clear message, easy to follow
Engagement 5 Hook is weak, "here's what happened" is overused
Authenticity 6 Generic advice, needs specific example
Purpose 7 Makes point but no discussion prompt

Overall: 6.4 - Needs work on hook and authenticity.

Draft 2:

My to-do list app sent me a "you haven't logged in for 180 days" email.

I smiled.

I was a to-do list addict. Todoist, Things, Notion databases with 47 properties. Then I noticed: my most productive days were when I forgot to check the list.

Now I use sticky notes. Physical ones. They fall behind my desk and disappear. That's the feature, not a bug. If something mattered, I would have done it already.

Controversial take: most "productivity" is just elaborate procrastination.

What's your relationship with to-do lists? Healthy or complicated?

Self-Critique:

Dimension Score Notes
Voice Match 8 Better casual tone, includes humor
Clarity 9 Very clear, builds logically
Engagement 9 Strong hook, curiosity created
Authenticity 8 Specific detail (47 properties), honest
Purpose 9 Ends with discussion prompt

Overall: 8.6 - Above threshold, ready to publish.

Configuration

Set Quality Threshold

Default threshold is 8.0. Adjust per task:

/critique-write blog post about X --threshold 9

Set Max Iterations

Prevent infinite loops:

/critique-write email about Y --max-rewrites 5

Focus on Specific Dimensions

/critique-write tweet about Z --focus engagement,voice

Voice Profile Integration

This skill works best with an established voice profile. If you don't have one:

  1. Run /learn-voice first with sample content
  2. Or provide examples in your request

Critique Explanations

Each critique includes:

  • What's working - Don't lose these elements
  • What's weak - Specific problems identified
  • How to fix - Actionable rewrite guidance

Advanced: Custom Critique Dimensions

For specialized content, add dimensions:

/critique-write technical tutorial --add-dimension "Accuracy: Are all code examples correct?"

When to Override

Sometimes you want draft 1 even if it scores low:

  • Time pressure
  • Draft is intentionally different
  • You'll edit heavily anyway

Just say: "Use draft 1, I'll edit it myself"

Credits

Inspired by @koylanai's "Ralph Wiggum Copywriter" - the insight that self-critique loops produce dramatically better output than single-pass generation.