Writing

Em Dash Remover

Scans any file for em dashes and replaces them with natural alternatives. Removes one of the most obvious AI writing tells from your copy in a single pass.

/humanizer Free
X-Ray: what this skill can and cannot do
Shell access No
Network calls No
File writes Yes (replaces em dashes in-place)
File reads User-specified files only
Destructive ops No
Credential access No
Scope Project only

Why this exists

Em dashes are one of the strongest tells that text was generated by AI. Humans rarely use them in casual or professional writing. AI models use them constantly, sometimes three or four times per paragraph.

If you are using Claude to draft copy, proposals, documentation, or marketing content, those em dashes make it obvious. This skill strips them out in a single pass and replaces each one with the punctuation a human writer would actually reach for.

How it works

Specify File
Find Em Dashes
Pick Replacement
Report Changes
  1. You specify the file(s) Pass a file path or glob pattern as an argument. The skill will scan every matching file.
  2. It finds all em dashes Searches for both spaced () and unspaced () variants.
  3. It picks the right replacement Each em dash is replaced based on the surrounding context: comma for a mid-sentence pause, colon before a result or list, full stop to split clauses, brackets for an aside.
  4. It reports what changed After processing, you get a count of how many replacements were made per file.

Step by step (for first-time users)

  1. Write your content using Claude Blog post, proposal, email, whatever. Save the file when you are done.
  2. Open Claude Code in that project Navigate to the folder where your file lives and launch Claude Code.
  3. Type /humanizer path/to/your-file.md Replace the path with your actual file. If you skip the path, Claude will ask which file to scan.
  4. Claude scans the file and finds every em dash It reads the surrounding sentence for each one to decide the right replacement.
  5. Each em dash is replaced with natural punctuation Commas, colons, full stops, or brackets, depending on context.
  6. Done. The file is updated. Check the diff if you want, but it will look natural. Commit when happy.

Replacement rules

ContextReplace withExample
Mid-sentence pauseCommaBrand strategy, and how it applies
Introducing a result or listColonThe result: a brand that works
Two related clausesFull stopClean and minimal. Always intentional.
Parenthetical asideBracketsStudio (founded 2004)

Honest take

What it does well: One job, done fast. It finds every em dash in your file and replaces it with the punctuation a human would actually use. It does not swap blindly. It reads the surrounding sentence and picks between a comma, colon, full stop, or brackets depending on context. I built this because I was tired of editing the same AI tell out of every piece of copy I generated. Now I run it once and move on.

What it does not do: Em dashes are not the only AI tell. Words like "delve," "leverage," "it's worth noting," the overuse of lists, the way every paragraph sounds like a LinkedIn post. This skill does not touch any of that. It also does not restructure sentences. If Claude used an em dash in a place where the whole sentence needs rewriting, the replacement will read slightly off. Give it a quick scan after the pass. Most of the time, it is clean.

When to use it: After any session where Claude helped you write. Blog posts, proposals, client emails, landing pages, documentation. Anything that a reader might look at and think "a machine wrote this." Run it before you publish, not after someone points it out.

Use this skill in your project

Download the .md file, drop it into .claude/skills/, and run /humanizer.

Download .md
Share
X LinkedIn Reddit