跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
效率与协作

document-superpowers

Complete document writing workflow with four stages: brainstorming → planning → execution → review. Use when user wants to write articles, blog posts, technical docs, or any long-form content.

binbinao · 已有 AI 审核记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add binbinao/document-superpowers --skill document-superpowers

技能说明


name: document-superpowers description: "Complete document writing workflow with four stages: brainstorming → planning → execution → review. Use when user wants to write articles, blog posts, technical docs, or any long-form content."

Document Superpowers

A systematic writing workflow that produces professional-quality articles, blog posts, and documentation.

When to Use This

Trigger this skill when the user wants to:

  • Write a blog post
  • Create technical documentation
  • Draft an article or essay
  • Produce any long-form content (>500 words)
  • Write something that needs structure and quality

Key phrases that should trigger:

  • "Write a blog post about..."
  • "I need to create documentation for..."
  • "Help me write an article..."
  • "I want to write about..."

Overview

Document Superpowers enforces a four-stage writing process:

💡 Brainstorming → 📋 Planning → ✍️ Execution → 🔍 Review

Each stage has strict gates - you cannot skip ahead without completing the previous stage.

The Four Stages

Stage 1: Brainstorming

Skill: document-superpowers:brainstorming

What it does:

  • Asks clarifying questions one at a time
  • Understands audience, purpose, core message
  • Presents topic brief for approval
  • Saves approved brief

Output: docs/writing/YYYY-MM-DD-<topic>-brief.md

Next: Automatically invokes writing-article-plan

Stage 2: Planning

Skill: document-superpowers:writing-article-plan

What it does:

  • Creates detailed section-by-section outline
  • Specifies key points, evidence needs, word counts
  • Plans transitions between sections
  • Identifies research requirements upfront

Output: docs/writing/YYYY-MM-DD-<topic>-plan.md

Next: Offers choice of execution mode

Stage 3: Execution

Skills:

  • document-superpowers:executing-article-plan (batch mode)
  • document-superpowers:subagent-driven-writing (iterative mode)

What it does:

  • Writes sections according to plan
  • Self-checks each section
  • Provides checkpoints for human review
  • Merges into full draft

Output:

  • Section drafts: docs/writing/drafts/YYYY-MM-DD-<topic>/section-N-*.md
  • Full draft: docs/writing/YYYY-MM-DD-<topic>-draft.md

Next: Automatically invokes reviewing-article

Stage 4: Review

Skill: document-superpowers:reviewing-article

What it does:

  • Four-pass systematic review:
    1. Logic (is it sound?)
    2. Evidence (is it supported?)
    3. Flow (is it smooth?)
    4. Polish (is it clean?)
  • Categorizes issues by severity
  • Documents all fixes
  • Produces final article

Output:

  • Review log: docs/writing/YYYY-MM-DD-<topic>-review.md
  • Final article: docs/writing/YYYY-MM-DD-<topic>-final.md

Quick Start

For Users

  1. Say: "I want to write about [topic]"
  2. Answer brainstorming questions
  3. Approve the brief
  4. Approve the plan
  5. Choose execution mode (batch or iterative)
  6. Review checkpoints as writing progresses
  7. Approve final article

For Agents

  1. Detect writing intent - Look for phrases like "write", "create article", "blog post"
  2. Invoke brainstorming - Start with questions, never skip to writing
  3. Follow the chain - Each skill invokes the next automatically
  4. Enforce gates - Do not skip stages or write before planning
  5. Provide checkpoints - Regular human validation points

File Structure

A complete writing project creates:

docs/writing/
├── YYYY-MM-DD-topic-brief.md          # Stage 1 output
├── YYYY-MM-DD-topic-plan.md           # Stage 2 output
├── drafts/YYYY-MM-DD-topic/           # Stage 3 working directory
│   ├── section-1-intro.md
│   ├── section-2-argument.md
│   └── ...
├── YYYY-MM-DD-topic-draft.md          # Stage 3 output
├── YYYY-MM-DD-topic-review.md         # Stage 4 review log
└── YYYY-MM-DD-topic-final.md          # Stage 4 final output

Execution Modes

Batch Execution (Default)

  • Write 2-3 sections at a time
  • Checkpoint after each batch
  • Fewer interruptions, longer focus
  • Good for: Straightforward articles, experienced writers

Subagent-Driven (Advanced)

  • Fresh subagent per section
  • Two-stage review (spec + quality)
  • Tight feedback loop
  • Good for: Complex articles, high-stakes content, learning the system

Key Principles

1. No Skipping Brainstorming

Every article goes through all four stages. Even "simple" topics need briefs (they can be short, but they must exist and be approved).

2. One Question at a Time

Don't overwhelm users with forms or multiple questions. Build understanding through conversation.

3. Bite-Sized Sections

Sections are 150-400 words - small enough to write in one sitting, large enough to develop an idea.

4. Checkpoints Not Marathons

Review after every 2-3 sections. Catch issues early.

5. Four-Pass Review

Each review pass focuses on one dimension. Don't mix logic checking with grammar checking.

Anti-Patterns to Avoid

❌ "This is too simple to need a brief"

  • Every article needs a brief, even if it's three sentences

❌ Writing before planning

  • Always create the plan before executing

❌ Skipping self-checks

  • Validate each section before moving on

❌ Marathon writing sessions

  • Use checkpoints every 2-3 sections

❌ Single-pass review

  • Four passes catch more issues than one combined pass

Example Usage

See examples/typescript-blog-post.md for a complete walkthrough.

Quick example:

User: "I want to write a blog post about code reviews"

Agent: [invokes brainstorming]
"Who's your target audience?"

User: "Junior developers"

Agent: "What should they do after reading?"

User: "Write better code review comments"

Agent: [continues questions → presents brief → gets approval]
Agent: [creates plan → offers execution choice]

User: "Batch execution"

Agent: [writes sections 1-3]
Agent: "Checkpoint: sections 1-3 complete. Continue?"

User: "Yes"

Agent: [completes draft → runs review → fixes issues]
Agent: "Article compl
阅读完整来源 ↗