AgentSkillsCN

Review Readme

进阶

SKILL.md

Review README.md

Review the Blog in a Prompt README.md for factual accuracy and clarity for non-technical users.

Trigger phrases

  • "review the readme"
  • "check the readme"
  • "audit README.md"
  • "is the readme clear"

What this skill does

Performs a comprehensive review of README.md from two perspectives: technical accuracy and non-technical user clarity.

Part 1: Factual accuracy

1.1 URL verification

  • Check that all URLs are valid and point to the correct resources
  • Verify GitHub raw URLs match the actual repo name
  • Check that external links (nodejs.org, github.com, etc.) are correct

1.2 Command accuracy

  • Verify all shell commands are syntactically correct
  • Check that Mac commands work in Terminal/zsh
  • Check that Windows commands work in PowerShell
  • Verify installation commands use current/correct package names

1.3 Consistency with PROMPT.md

  • Verify claims about what the blog includes/excludes match PROMPT.md
  • Check that the publishing workflow description matches the actual flow
  • Ensure any technical details align with the implementation

1.4 Version/tool accuracy

  • Verify referenced tools exist and work as described
  • Check that installation methods are current (e.g., Claude Code install command)
  • Flag any outdated information

Part 2: Non-technical user clarity

2.1 Jargon detection

  • Flag technical terms that aren't explained (git, repo, CLI, terminal, etc.)
  • Identify acronyms without definitions
  • Check that explanations use simple analogies where helpful

2.2 Assumption audit

  • Identify places that assume prior knowledge
  • Flag steps that might be confusing to someone who's never used Terminal/PowerShell
  • Check for missing context that a beginner would need

2.3 Instruction clarity

  • Verify each step is actionable and specific
  • Check that users know exactly what to type/click
  • Ensure success criteria are clear (how do they know it worked?)

2.4 Error handling guidance

  • Check if common failure scenarios are addressed
  • Verify troubleshooting advice is accessible to non-technical users
  • Ensure error recovery doesn't require technical knowledge

2.5 Visual/formatting clarity

  • Check that code blocks are clearly marked
  • Verify the distinction between "type this" vs "you'll see this"
  • Ensure important warnings stand out

Part 3: User journey analysis

3.1 Complete path verification

  • Walk through the entire setup as a new user would
  • Identify any gaps or missing steps
  • Check that transitions between sections are smooth

3.2 Cognitive load assessment

  • Flag sections that might overwhelm a beginner
  • Identify opportunities to break complex steps into smaller ones
  • Check that the most important information comes first

Output format

Provide findings in these categories:

Factual Issues

Issues where the README states something incorrect or outdated.

Clarity Issues

Places where a non-technical user might get confused or stuck.

Suggestions

Improvements that would enhance the user experience.

For each finding, include:

  1. The specific location (line number or section)
  2. The current text (if relevant)
  3. What the issue is
  4. Suggested improvement

Example output

code
## Factual Issues

### 1. Outdated installation command
- **Location**: Line 51, Windows Claude Code installation
- **Current**: `irm https://claude.ai/install.ps1 | iex`
- **Issue**: This URL may have changed; verify it's current
- **Suggested**: Confirm URL at claude.ai/download or update to current method

## Clarity Issues

### 1. "Terminal" not explained for beginners
- **Location**: Line 35, "Open Terminal"
- **Current**: "Open Terminal (Applications folder > Utilities folder > Terminal)"
- **Issue**: Users may not know what Terminal is or why they need it
- **Suggested**: Add brief explanation: "Terminal is an app that lets you type commands to your computer. Think of it as a way to talk directly to your Mac."

### 2. Unclear success indicator
- **Location**: Line 46, after `brew install node`
- **Issue**: User doesn't know what success looks like
- **Suggested**: Add: "When it's done, you'll see your prompt again (the line where you type). You can verify it worked by typing `node --version` - you should see a number like v20.x.x"

## Suggestions

### 1. Add estimated time for each step
- **Location**: Throughout "How to Create Your Blog" section
- **Suggestion**: Help users plan by adding "(~2 minutes)" after each major step