Research Writing Skill

A Claude Code skill for producing informed, measured, evidence-driven long-form prose that reads as thoughtful and credible rather than promotional or AI-generated.

Overview

This skill helps you write blog posts, research summaries, product announcements, technical narratives, industry analyses, white papers, thought leadership pieces, and any long-form content that should feel researched, authoritative, and human.

The skill follows a proven three-phase workflow and incorporates the humanizer skill to eliminate AI writing patterns and ensure natural, human-like prose.

Installation

1. Copy SKILL.md to your Claude Code skills directory (~/.claude/skills/research-writing/)
2. The skill will be automatically available in Claude Code

How to Use

Trigger Phrases

The skill automatically activates when you use phrases like:

• "write a blog post"
• "draft an announcement"
• "research-style writing"
• "make this sound informed"
• "write this like a real person"
• "thoughtful," "measured," or "credible" writing requests

Basic Usage

Simply provide a topic or brief and the skill will guide you through the process:

/research-writing "Write a blog post about our new API features"

Or just ask naturally:

"Can you help me write a research summary about machine learning performance improvements?"

The Three-Phase Workflow

Phase 1: Interview

The skill conducts a structured interview to gather crucial context:

• Audience and intent: Who is this for? What should they take away?
• Evidence and specifics: What data, benchmarks, quotes, or examples are available?
• Structure and constraints: What article type fits? Any length or style requirements?
• Edge cases and tradeoffs: What tensions or limitations should be addressed?

Phase 2: Draft

Creates the full draft following proven principles:

• Evidence-backed claims with specific numbers and comparisons
• Clear article structure (announcement, research report, narrative, or industry piece)
• Professional but not corporate tone
• Technical accuracy with accessible explanations

Phase 3: Humanizer Pass

Automatically invokes the humanizer skill to eliminate AI writing patterns:

• Removes significance inflation ("revolutionary," "game-changing")
• Fixes synonym cycling and elegant variation
• Eliminates promotional language and filler phrases
• Corrects structural tells and AI vocabulary patterns

Input Examples

Example 1: Product Announcement Brief

**Brief**: "We're launching a new code review feature that uses AI to catch bugs automatically. It reduced review time by 40% in our internal testing and found 23% more critical issues than human reviewers alone."

**Additional Context**:
- Target audience: Software engineering teams
- Key differentiator: Runs in real-time during development
- Available data: 6-month internal trial with 50 developers
- Publication: Company blog and developer newsletter

The skill will interview to gather:

• Specific benchmark comparisons
• User quotes from the trial
• Technical implementation details
• Pricing and availability information
• Limitations or caveats to mention

Example 2: Research Summary Brief

**Brief**: "Write a research summary about our study on remote work productivity. We surveyed 1,200 knowledge workers across 6 months and found mixed results depending on job type and experience level."

**Additional Context**:
- Target audience: HR leaders and business executives
- Key finding: Productivity varied significantly by role
- Available data: Survey responses, productivity metrics, interview transcripts
- Length: 2,500-3,000 words

The skill will interview to gather:

• Specific productivity metrics and statistical significance
• Methodology details for credibility
• Demographic breakdown of survey participants
• Conflicting or surprising findings to address
• Implications for policy recommendations

Example 3: Technical Narrative Brief

**Brief**: "Tell the story of how our engineering team solved the database scaling problem that was causing outages. We went from 12 outages per month to zero over 4 months."

**Additional Context**:
- Target audience: Technical leadership and engineering teams
- Story arc: Problem identification → Solution exploration → Implementation → Results
- Available sources: Engineering team interviews, incident reports, performance data

The skill will interview to gather:

• Technical details about the scaling solution
• Specific timeline and milestones
• Team member quotes about challenges faced
• Quantitative improvements beyond outage reduction
• Lessons learned for other engineering teams

Article Type Examples

Product Announcements

• Open with clear, concise framing
• Present capabilities with specific numbers
• Include benchmark comparisons
• Feature customer validation quotes
• End with availability details

Research Reports

• Frame with clear research questions
• Explain methodology accessibly
• Present findings with appropriate precision
• Acknowledge limitations and caveats
• Connect to broader implications

Technical Narratives

• Start with concrete problem context
• Build from specific to general insights
• Use detailed examples to illustrate concepts
• Include team perspectives and quotes
• Extract practical lessons for readers

Industry Analyses

• Frame around real practitioner problems
• Demonstrate concrete use cases
• Include specific partner outcomes
• Avoid generic industry predictions
• Focus on actionable insights

Key Writing Principles

Evidence Over Assertion

❌ "The system shows major improvements" ✅ "The new model scores 76% on the 8-needle retrieval benchmark, compared to 18.5% for the previous version"

Technical Terms in Context

❌ "The Gini coefficient fell. (Note: The Gini coefficient measures inequality.)" ✅ "The Gini coefficient, a standard measure of equality, fell from 0.37 to 0.32"

Intellectual Honesty

❌ "Our solution completely eliminates the problem" ✅ "The system reduces false positives by 60%, though it still lags behind human experts in edge cases"

What Gets Avoided

The skill actively eliminates common AI writing patterns:

• Hype inflation: "revolutionary," "game-changing," "unprecedented"
• Generic conclusions: "The future looks bright," "exciting times ahead"
• AI vocabulary: "delve," "leverage," "landscape" (abstract usage)
• Promotional language: "boasts," "showcasing," "vibrant"
• Structural tells: Excessive bullet points with inline headers
• Synonym cycling: Artificially rotating between "system," "platform," "solution"

Attribution

This skill incorporates and automatically invokes the humanizer skill created by @blader. The humanizer performs the final pass to eliminate AI writing patterns and ensure natural, human-like prose.

The research-writing methodology is based on analysis of effective technical and scientific communication across academic, industry, and journalistic contexts.

Example Output Quality

Before: "Our groundbreaking new platform leverages cutting-edge AI to revolutionize the development landscape, offering unprecedented capabilities that will transform how teams collaborate."

After: "The new code review tool catches 23% more critical bugs than manual reviews alone. In our six-month trial with 50 developers, it reduced review cycles from an average of 2.3 days to 1.4 days while maintaining code quality standards."

License

MIT License - See LICENSE file for details.

Contributing

Contributions welcome! Please:

1. Test changes with diverse writing briefs
2. Ensure the three-phase workflow remains intact
3. Maintain attribution to the humanizer skill
4. Include examples of before/after output quality

Support

For issues or questions:

• Open an issue in this repository
• Reference specific examples of unexpected behavior
• Include the input brief and desired output type