Skip to main content
skill CX Codex

remove-ai-style — Claude Code skill: strip AI writing patterns

Rewrites English or Chinese prose to drop formulaic phrasing, stray dashes and exclamation marks. Useful before you publish an article or a doc.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add zc277584121/marketing-skills --skill remove-ai-style -a claude-code
Open on GitHub

What remove-ai-style does

The skill runs a two-layer pass. A deterministic Markdown analyzer goes first and flags repeatable signals: word choice, punctuation, structure, leftover assistant phrasing. Then the agent reads the whole piece and fixes what no rule list can catch — symmetrical argument, paragraphs that repeat the same move in fresh words, empty evaluative adjectives, sentences that explain the point to the reader.

Three rule families are hard constraints on editable prose. Every exclamation mark goes. So do em dashes, en dashes and double hyphens standing in for a dash, along with formulaic binary contrasts such as “not just X but Y”. After the rewrite, rerunning the analyzer with --fail-on-hard-constraints should report zero for those counters. Code, URLs, Markdown targets and quoted source that must stay verbatim are the exception.

You need python3 and a path to the article, or the text piped through stdin. The analyzer lives at scripts/analyze_ai_style.py inside the skill folder, and swapping it for hand-rolled grep lines defeats the purpose — detection logic sits in that script. Before editing, the skill records what has to survive: YAML frontmatter, fenced code, tables, images and links, HTML comments, full VISUAL_TODO blocks. Numbers, facts, URLs and citation targets stay as they are, even when a smoother sentence is tempting.

What it helps with

  • clear exclamation marks and dashes before publishing
  • turn formulaic binary contrasts into direct claims
  • spot repeated paragraph logic hiding behind new vocabulary
  • cut empty evaluative adjectives and decorative transitions
  • check a generated article for assistant-style residue

Example prompts for the agent

  • Run the analyzer on article.md and show me the JSON report
  • Rewrite the flagged spots in draft.md, keep the facts and structure
  • Rescan with --fail-on-hard-constraints and list what is left

How to install remove-ai-style

1Skills CLI (Vercel)

npx skills add zc277584121/marketing-skills --skill remove-ai-style -a claude-code

2Manually: put the files in place

~/.claude/skills/remove-ai-style/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/remove-ai-style/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about remove-ai-style

How do I install remove-ai-style in Claude Code?

Run npx skills add zc277584121/marketing-skills --skill remove-ai-style -a claude-code. The skill lands at ~/.claude/skills/remove-ai-style/SKILL.md.

What does remove-ai-style do?

It finds formulaic phrasing and generation traces in English or Chinese text, then rewrites them. A deterministic Markdown pass comes first, followed by a semantic edit over the full article.

What do I need to run the analyzer?

python3 plus a path to the article, or text passed through stdin. The script sits at scripts/analyze_ai_style.py inside the skill folder.

Similar skills

All skills for Writing and docs →

From the same set4

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor