AgentSkillsCN

personal-docs-editor

审核并编辑文档,去除 AI 自动生成的企业化语言、营销式表述以及过度正式的措辞,将文档风格转变为轻松自然的个人化语气,同时完整保留技术性信息。

SKILL.md
--- frontmatter
name: personal-docs-editor
description: Reviews and edits documentation to remove AI-generated corporate language, marketing speak, and excessive formality, transforming docs into casual personal tone while preserving technical information.
version: 1.0.0
author: jdwork
category: documentation
tags: [editing, tone, cleanup]

Skill: Personal Documentation Editor

Description

Reviews and edits documentation in personal projects (like dotfiles) to remove "AI slop" - overly formal, corporate, or prompt-responsive language. Transforms documentation into a casual, personal tone while preserving all useful technical information.

Instructions

1. Identify AI Slop Patterns

Search for these red flags that indicate AI-generated corporate language:

Tone & Style Issues:

  • Marketing language: "fully-configured", "seamless integration", "comprehensive solution"
  • Corporate formality: "SECURITY:", "Best Practices", "Getting Started" (when simpler works)
  • Prompt-responsive phrases: "designed to be simple, modular, and easy to extend"
  • Excessive formality: "Please ensure that...", "It is recommended that..."
  • FAQ sections that answer questions nobody asked
  • Governance language: "Security Notes", "Compliance", "Best Practices"

Formatting Issues:

  • Excessive emoji in section headers (1-2 per doc is fine, 10+ is too much)
  • Bullet points with checkmarks (✅) - feels like marketing copy
  • Overly nested numbered lists (### 1. Configure, ### 2. Setup, etc.)
  • Table of Contents for docs under 200 lines
  • ASCII diagrams that are for show (keep genuinely useful ones)

Content Issues:

  • Explanatory sections that state the obvious
  • Security notes that are just governance theater
  • Extensive "Getting Help" sections for personal projects
  • Over-structured TODO lists with checkbox hierarchies
  • Contradictory or confusing comments

2. Assess Documentation Context

Before editing, determine:

  • Audience: Personal use or for AI agents? (AGENTS.md should stay formal)
  • Purpose: Reference docs can be detailed; quick starts should be brief
  • Existing tone: Check main README for voice/style to match
  • Technical depth: Keep all useful commands, troubleshooting, and details

3. Apply Personal Tone Transformations

Replace marketing copy with direct statements:

  • ❌ "A fully-configured, reproducible development environment"
  • ✅ "Docker container with Opencode and my dotfiles"

Simplify formal language:

  • ❌ "SECURITY: Your API keys are stored securely..."
  • ✅ "API keys live in ~/.env on your host (not in git)"

Casual over corporate:

  • ❌ "Optional but Recommended"
  • ✅ "optional but nice"
  • ❌ "Prerequisites" → "What you need"
  • ❌ "Additional Resources" → "Resources"

Simplify FAQ format:

  • ❌ "### Q: Can I use VS Code?\nA: Yes! Install the..."
  • ✅ "VS Code? Yes, install the..."

Remove unnecessary formality:

  • ❌ "### 1. Configure API Keys"
  • ✅ "### Configure API Keys"

Fix confusing comments:

  • ❌ "Password: use your SSH key (no password needed)"
  • ✅ "Uses your SSH key (no password needed)"

4. Preserve What's Useful

Keep these elements:

  • ✅ All technical details, commands, and troubleshooting
  • ✅ Diagrams that actually explain structure (ASCII art is fine if useful)
  • ✅ Detailed instructions and step-by-step guides
  • ✅ Code examples and configuration samples
  • ✅ FAQ content (just make it less formal)
  • ✅ Tables of keyboard shortcuts or command references
  • ✅ Table of contents for genuinely long docs (200+ lines)

5. Maintain Consistency

Ensure tone matches across all docs in the project:

  • Check main README for the project's voice
  • Personal projects: casual, first-person okay ("my setup", "I use")
  • Configuration for tools: straightforward, no fluff
  • Comments in code: minimal, purposeful, clear

6. Execute the Edits

Workflow:

  1. Scan: Find all .md, README, and heavily-commented config files
  2. Review: Read through and identify AI slop patterns
  3. Report: List findings by file with severity (High/Medium/Low priority)
  4. Confirm: Ask user which files to edit and what to preserve
  5. Edit: Make changes systematically, preserving technical content
  6. Summary: Report what was changed and why

Key principles:

  • Never remove useful technical information
  • Keep diagrams if they explain something
  • Simplify tone without losing clarity
  • Remove emoji spam but 1-2 for visual scanning is okay
  • Make it sound like a human wrote notes for themselves

Examples

Example 1: Marketing Language

Before:

markdown
## 🚀 Features

- ✅ Fully-configured development environment
- ✅ Seamless terminal integration
- ✅ Enterprise-grade security practices

After:

markdown
## Features

Docker container with my dev tools, SSH access, and API keys stored outside git.

Example 2: Formal FAQ

Before:

markdown
### Q: How much disk space does this use?

**A:** The base image requires approximately 2GB of disk space.

After:

markdown
**Disk space?** Base image is ~2GB, plus your projects.

Example 3: Security Theater

Before:

markdown
## 🔒 Security Notes

- API keys stored on separate device following best practices
- SSH uses key-based authentication only (no passwords)
- Container runs as non-root user for enhanced security
- Read-only mounts prevent accidental modification

After:

markdown
API keys live in ~/.env on your host (not in git). SSH uses keys, not passwords.

Example 4: Prompt-Responsive Language

Before:

markdown
This configuration is designed to be simple, modular, and easy to extend,
using lazy.nvim for efficient plugin management.

After:

markdown
My Neovim setup using lazy.nvim for plugin management.

Detection Heuristics

Use these patterns to automatically flag AI slop:

Regex patterns to search for:

  • designed to be \w+, \w+, and \w+
  • fully[- ]configured
  • seamless(ly)? integrat(e|ion)
  • best practices?
  • ensure that
  • please note
  • it is (recommended|important|advised)
  • \*\*SECURITY:\*\*
  • Multiple consecutive emoji headers: ^## [🎯🚀📦🛠️🔧]

Structural patterns:

  • 5+ emoji in headers within same document
  • Sections named "Getting Help" in personal projects
  • "Prerequisites" sections with nested bold headers
  • FAQ with "Q:" and "A:" format
  • TODO lists with ### 1., ### 2. numbering

Prerequisites

  • None - works on any documentation
  • Most effective on personal projects (dotfiles, tools, scripts)
  • Can be adapted for team docs by adjusting tone guidelines

Notes

When NOT to use this skill:

  • Documentation meant for external users (keep it clear and helpful)
  • Agent instructions (AGENTS.md should stay formal and precise)
  • Technical specifications that need formal language
  • When user wants to keep corporate style

Remember:

  • Personal documentation is for YOU - make it useful, not impressive
  • Clear > formal
  • Direct > marketing
  • Casual > corporate
  • Keep what helps, cut what shows off

Balance:

  • Don't remove politeness, just formality
  • Don't remove structure, just over-structure
  • Don't remove detail, just fluff
  • Don't remove emoji entirely, just excessive use