Document Superpowers gives your agent a fixed writing process instead of starting from a blank page. It uses separate skills for brainstorming, outlining, drafting, and review, with checkpoints between stages so you can correct the brief before the draft gets too far.
Builders who want their agent to write with a clear brief, a structured outline, and a review pass before publishing.
You can get structured drafts that are easier to review, revise, and publish.
Guides one-question-at-a-time topic exploration and turns it into an approved brief.
Creates a section-by-section outline with word counts, evidence needs, and transitions.
Writes the draft in batches with checkpoints after every few sections.
Runs a four-pass review for logic, evidence, flow, and polish, then records fixes.
Uses a fresh subagent per section for faster iteration on more complex pieces.
# Clone to Claude Code global skills directory git clone https://github.com/binbinao/document-superpowers.git ~/.claude/skills/document-superpowers
# Clone to Cursor global skills directory git clone https://github.com/binbinao/document-superpowers.git ~/.cursor/skills/document-superpowers # Or clone to project-level directory git clone https://github.com/binbinao/document-superpowers.git .cursor/skills/document-superpowers
# Clone to Codebuddy global skills directory git clone https://github.com/binbinao/document-superpowers.git ~/.codebuddy/skills/document-superpowers # Or clone to project-level directory git clone https://github.com/binbinao/document-superpowers.git .codebuddy/skills/document-superpowers
# Clone to Gemini CLI config directory git clone https://github.com/binbinao/document-superpowers.git ~/.gemini/skills/document-superpowers
# Option 1: Clone to your workspace skills directory git clone https://github.com/binbinao/document-superpowers.git ~/.openclaw/skills/document-superpowers # Option 2: Install via OpenClaw skills manager (if available) openclaw skills install document-superpowers
"I want to write about [topic]"
📄 Deep-dive case study with metrics, highlights, and architecture: binbinao.github.io/resume/projects/document-superpowers/
A complete document writing workflow for your AI writing agents, inspired by obra/superpowers code development methodology.
Document Superpowers transforms your AI from a "just write something" tool into a systematic writing partner. It enforces a proven four-stage process that produces higher-quality articles, blog posts, and documentation.
When you ask your AI to write something, it doesn't just start typing. Instead:
Brainstorming - It steps back and asks what you're really trying to communicate. Who's the audience? What's the core message? What should readers do after reading?
Planning - Once it understands the topic, it creates a detailed outline breaking the article into manageable sections (150-400 words each), complete with key points, evidence needs, and transitions.
Execution - It writes section by section, with self-checks and validation at each step. You get checkpoints every 2-3 sections to review progress.
Review - Systematic four-pass review covering logic, evidence, flow, and polish. Every issue gets categorized, documented, and fixed.
The result? Professional-quality writing that's actually readable, well-structured, and achieves its purpose.
💡 Brainstorming → 📋 Planning → ✍️ Execution → 🔍 Review
Each stage has strict gates - you can't skip ahead without completing the previous stage. No more "I'll just wing it" articles.
# Clone to Claude Code global skills directory
git clone https://github.com/binbinao/document-superpowers.git ~/.claude/skills/document-superpowers
After cloning, add skill references in ~/.claude/CLAUDE.md or your project's .claude/CLAUDE.md:
## Skills
- ~/.claude/skills/document-superpowers/skills/brainstorming/SKILL.md
- ~/.claude/skills/document-superpowers/skills/writing-article-plan/SKILL.md
- ~/.claude/skills/document-superpowers/skills/executing-article-plan/SKILL.md
- ~/.claude/skills/document-superpowers/skills/reviewing-article/SKILL.md
- ~/.claude/skills/document-superpowers/skills/subagent-driven-writing/SKILL.md
# Clone to Cursor global skills directory
git clone https://github.com/binbinao/document-superpowers.git ~/.cursor/skills/document-superpowers
# Or clone to project-level directory
git clone https://github.com/binbinao/document-superpowers.git .cursor/skills/document-superpowers
After cloning, add the skill paths in Cursor's Rules or Agent Settings.
# Clone to Codebuddy global skills directory
git clone https://github.com/binbinao/document-superpowers.git ~/.codebuddy/skills/document-superpowers
# Or clone to project-level directory
git clone https://github.com/binbinao/document-superpowers.git .codebuddy/skills/document-superpowers
After cloning, add the skill paths in Codebuddy's configuration.
# Clone to Gemini CLI config directory
git clone https://github.com/binbinao/document-superpowers.git ~/.gemini/skills/document-superpowers
Add skill references in ~/.gemini/GEMINI.md or your project's GEMINI.md:
## Skills
- ~/.gemini/skills/document-superpowers/skills/brainstorming/SKILL.md
- ~/.gemini/skills/document-superpowers/skills/writing-article-plan/SKILL.md
- ~/.gemini/skills/document-superpowers/skills/executing-article-plan/SKILL.md
- ~/.gemini/skills/document-superpowers/skills/reviewing-article/SKILL.md
- ~/.gemini/skills/document-superpowers/skills/subagent-driven-writing/SKILL.md
# Option 1: Clone to your workspace skills directory
git clone https://github.com/binbinao/document-superpowers.git ~/.openclaw/skills/document-superpowers
# Option 2: Install via OpenClaw skills manager (if available)
openclaw skills install document-superpowers
Copy the skills/ directory into your AI agent's skills folder according to your platform's documentation.
"I want to write about [topic]"
Your agent will automatically invoke the brainstorming skill and start asking questions.
The agent asks one question at a time to understand:
Once it understands your topic, it presents a brief covering:
Approve or request revisions.
After brief approval, it creates a detailed outline:
Batch execution (default):
Subagent-driven (advanced):
Systematic four-pass review:
Issues get categorized (critical/moderate/minor) and fixed systematically.
Your final article is saved with:
User: "I want to write a blog post about why engineers should write documentation"
Agent: [invokes brainstorming]
"Let me understand what you're building. Who's your target audience - junior engineers, senior engineers, or engineering managers?"
User: "Senior engineers who hate writing docs"
Agent: "Got it. What do you want them to do after reading - start writing better docs, advocate for documentation culture, or something else?"
User: "Change their mind that documentation is valuable"
Agent: [continues questions...]
Agent: [presents brief...]
Agent: [gets approval...]
Agent: [creates plan...]
Agent: [offers execution choice...]
User: "Batch execution"
Agent: [writes sections 1-3...]
Agent: "Checkpoint: Sections 1-3 complete (612 words). Continue?"
User: "Yes"
Agent: [writes sections 4-5...]
Agent: "Draft complete! Ready for review?"
User: "Yes"
Agent: [runs four-pass review...]
Agent: "Review complete. 2 critical issues found and fixed. Final article ready!"
Additional documentation is available in the docs/ directory:
docs/
├── DEVELOPMENT.md # Development overview & architecture details
├── SUMMARY.md # Implementation summary & project statistics
└── examples/
└── typescript-blog-post.md # Complete end-to-end writing walkthrough
| Document | Description |
|---|---|
| DEVELOPMENT.md | In-depth development guide covering project architecture, skill design, workflow details, and extension points |
| SUMMARY.md | Implementation summary with project statistics, file inventory, and feature highlights |
| typescript-blog-post.md | A complete example walkthrough: writing a TypeScript blog post from brainstorming to final review |
During a writing project, the workflow generates the following files:
docs/writing/
├── YYYY-MM-DD-topic-brief.md # Approved topic brief
├── YYYY-MM-DD-topic-plan.md # Detailed outline
├── drafts/
│ └── YYYY-MM-DD-topic/
│ ├── section-1-intro.md # Individual section drafts
│ ├── section-2-argument.md
│ └── ...
├── YYYY-MM-DD-topic-draft.md # Merged full draft
├── YYYY-MM-DD-topic-review.md # Review log
└── YYYY-MM-DD-topic-final.md # Published version
Everything is version-controlled and traceable.
| Code (superpowers) | Content (document-superpowers) |
|---|---|
| brainstorming → design doc | brainstorming → topic brief |
| writing-plans → implementation plan | writing-article-plan → article outline |
| executing-plans → code | executing-article-plan → draft |
| requesting-code-review | reviewing-article |
| TDD (red-green-refactor) | write-check-revise |
| Git worktrees | Section drafts |
| Subagent per task | Subagent per section |
Every article goes through the full process, even "simple" blog posts. Simple topics can have short briefs (a few sentences), but you MUST present and approve the brief before writing.
The agent never overwhelms you with multiple questions. Brainstorming is a conversation, not a form to fill out.
Sections are 150-400 words - small enough to write in one sitting (15-30 minutes), large enough to develop an idea.
Review after every 2-3 sections. Catch issues early rather than discovering problems in a 3000-word wall of text.
Each pass focuses on one dimension:
Mixing concerns makes issues harder to spot.
Every stage produces artifacts:
Trust the process - It feels slower at first, but produces better results faster than iterative rewrites.
Be specific in brainstorming - The clearer your answers, the better the brief.
Review checkpoints seriously - Don't just rubber-stamp. Actually read the sections.
Save the review log - It's valuable documentation of your writing decisions.
Enforce the gates - Never skip brainstorming, even for "simple" articles.
Ask better questions - Multiple choice > open-ended when possible.
Self-check rigorously - Catch your own issues before human review.
Preserve what works - Don't over-edit during review. Fix problems, keep strengths.
Create templates for specific document types:
skills/brainstorming/templates/
├── blog-post.md
├── technical-guide.md
├── api-documentation.md
└── research-paper.md
Add organization-specific style guidelines:
skills/reviewing-article/style-guides/
├── ap-style.md
├── company-voice.md
└── technical-writing.md
Use with other skills:
Skills live directly in this repository. To contribute:
skills/writing-skills/ in obra/superpowers for guidelines)Agent skips brainstorming:
Sections too long/short:
Review finds nothing:
Subagent fails repeatedly:
MIT License - see LICENSE file for details
This project is heavily inspired by obra/superpowers by Jesse Vincent. The systematic, gate-driven approach from code development translates beautifully to content creation.
Happy writing! 📝
Sign in to join the discussion.
No comments yet. Be the first to say what this is good for.
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.
Topic in, narrated explainer video out. A Claude Code / Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles and a chapter progress bar. Chinese or English; every frame drawn in code with Remotion.
Public repository for Agent Skills
Open-source AI job search: scan job portals, evaluate listings into a structured A-H report with a global 1-5 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
A skill to stop your coding agent from burying the answer. ADHD-friendly output.