claude-code-public

内容来源:README.md(说明文档) · 原始地址 · 查看安装指南

原始内容

Claude Code Agents & Skills

A collection of custom agents and skills (slash commands) for Claude Code, Anthropic's official CLI tool for Claude.

What's Included

Agents

Agent Description
newsletter-digest Process newsletters from Gmail into summarized vault digests. Fetches emails, filters spam/promos, extracts insights, and generates clean markdown summaries.
qa-audit-analytics QA auditor for data analytics projects. Verifies calculations, cross-references claims against source data, and flags discrepancies before outputs go into presentations or reports.

Skills (Slash Commands)

Skill Description
audit-sensitive Audit repository for sensitive information before pushing. Checks for emails, credentials, client names, PII, and other data that shouldn't be public.
humanize Transform AI-generated content into natural, human-sounding text. Identifies and fixes AI patterns like overused words, uniform sentence lengths, and robotic phrasing.

Installation

Prerequisites

Installing Agents

Copy agent files to your Claude Code agents directory:

# Create agents directory if it doesn't exist
mkdir -p ~/.claude/agents

# Copy agent file(s)
cp agents/qa-audit-analytics.md ~/.claude/agents/

Agents are automatically invoked by Claude Code when relevant tasks are detected. You can also reference them explicitly in your prompts.

Installing Skills (Commands)

Copy command files to your Claude Code commands directory:

# Create commands directory if it doesn't exist
mkdir -p ~/.claude/commands

# Copy command file(s)
cp commands/audit-sensitive.md ~/.claude/commands/
cp commands/humanize.md ~/.claude/commands/

Once installed, invoke skills using the slash command syntax:

/humanize

Usage Examples

QA Audit Analytics Agent

The QA audit agent is automatically invoked when you have a data analysis project that needs verification. You can also explicitly request an audit:

Audit the analysis in results.md for accuracy before I send it to the client.

The agent will:

  1. Discover all analysis files (scripts, data, reports)
  2. Extract numerical claims from output files
  3. Trace each claim back to source data
  4. Verify calculations, percentages, and counts
  5. Generate a discrepancy report with severity levels

Audit Sensitive Skill

Use the /audit-sensitive command before pushing to any remote:

/audit-sensitive

The skill scans your repository for:

  • Credentials and API keys
  • Email addresses and PII
  • Client/company names
  • File paths with usernames
  • Sensitive data in git history

Run this before every push to ensure nothing sensitive leaks.

Humanize Skill

Use the /humanize command to transform AI-generated text:

/humanize

[Paste your AI-generated content here]

The skill analyzes text for AI patterns and rewrites it to sound more natural by:

  • Varying sentence lengths (adding short punchy sentences)
  • Replacing flagged vocabulary (delve, tapestry, leverage, etc.)
  • Adding human elements (contractions, specifics, opinions)
  • Breaking predictable structures

Contributing

Feel free to submit issues or pull requests with improvements to existing agents/skills or suggestions for new ones.

Resources

License

MIT License - feel free to use and modify these agents and skills for your own projects.