Skip to content
SkillModel-led7,344 tokens in use

humanizer

Rewrites AI-sounding text so it reads like a person wrote it: it marks 26 known tells, drafts a rewrite and checks that every claim survives.

71,585 stars on the repository MIT Licenseblader/humanizer
/plugin marketplace add blader/humanizer /plugin install humanizer@humanizer
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing humanizer costs

72tokens idle
7,344tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works. Scripts: scripts/validate-package.py.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts and code blocks. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter15/15
  • Structure25/25
  • Examples6/15
  • Output15/15
  • Size13/15
  • Resources0/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Excellent
  • Development activity100
  • Maintenance96
  • Community64
  • Maturity32
  • Momentum40

Humanizer is an editing skill for prose that sounds machine-written. It works from Wikipedia’s “Signs of AI writing” guide: it reads the whole text, marks every tell it finds starting with the strongest, rewrites the passage and checks the draft against the original claims before it returns the final version.

It handles pasted text and whole files. For pasted text it shows the first rewrite, the tells that survived it and the final version. For a file it edits only the prose and leaves code, data, frontmatter and link targets alone. A writing sample makes the rewrite follow the writer’s own rhythm and punctuation.

What You Get

  • A /humanizer command that rewrites pasted text and returns the first rewrite, the tells that survived it and a final version
  • A file mode that reads a whole document from disk, edits only the prose and reports how many sections it reviewed
  • A catalogue of 26 numbered AI-writing patterns with before and after examples, ordered from strongest to weakest
  • Voice matching: give it two or three paragraphs of your writing and it follows their sentence length, word choice and punctuation

Common Use Cases

  • Cleaning up a blog post or newsletter draft that an AI wrote before it is published
  • Editing a pull request description or commit message so it reads like the author wrote it
  • Rewriting a long document in place while leaving its code blocks, links and metadata untouched
  • Reviewing a reply in a thread so it opens with the decision instead of re-explaining the background

What Makes It Unique

Most humanizer prompts swap a list of banned words. This skill orders its 26 patterns by how rarely a careful writer would use them on purpose, so five strong ones justify an edit on one sighting and weaker ones, such as dashes or curly quotes, need other tells nearby. It also protects what the text says: the draft is checked for lost names, numbers, dates and quotes, and informal human text such as chat messages is left alone. The README says it edits for readers and does not try to beat AI detectors.

Under The Hood

The skill is one SKILL.md of instructions with no scripts that run during use. When invoked, the agent reads the text, marks the patterns, drafts a rewrite, checks the draft against the original claims and writes the final version. In file mode it reads the file from disk and writes the edited prose back to it. The repository also holds a package validation script and plugin manifests for Claude Code, Cursor and pi, which belong to publishing and are not run by the skill.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers