Skip to content
Agentmart
Submit a tool

humanizer

Rewrite text so it reads like a person wrote it, without changing what it says.

Skill by ronnnerr

IndexedMIT★ 1
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/ronnnerr-humanizer

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Rewrite text so it reads like a person wrote it, without changing what it says. Run before delivering any human-facing prose: agency and client copy, landing pages, emails, video scripts, hooks, titles, descriptions, captions, READMEs, release notes, PR bodies. Use when someone says humanize this, this reads like AI, make it sound human, clean up the writing, or does this sound like a robot.

Use cases: writing, productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/humanizer

What's inside

Skill name
humanizer
Description
Rewrite text so it reads like a person wrote it, without changing what it says. Run before delivering any human-facing prose: agency and client copy, landing pages, emails, video scripts, hooks, titles, descriptions, captions, READMEs, release notes, PR bodies. Use when someone says humanize this, this reads like AI, make it sound human, clean up the writing, or does this sound like a robot.
License
MIT
Metadata
version: 1.0.0 · norstack: true
Instructions
279 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# humanizer

Strip the machine cadence out of a draft. Keep every claim.

The pattern taxonomy below is norstack's own, informed by Wikipedia's
[Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing)
(WikiProject AI Cleanup, CC BY-SA) and by StoryScope, arXiv 2604.03136, for the
narrative layer in §F.

## The one hard rule

The pass changes how a draft reads. It never changes what it says.

Do not add a fact, name, number, date, quote, citation, or ranking that is not in
the source. Do not drop a claim because it makes a sentence awkward. If a sentence
resists rewriting, restructure the paragraph around the claim. If a claim needs a
detail nobody supplied, ask for it or write the shorter true sentence.

Fiction is the only exemption. Invented detail is the job there.

## Modes

Pick by what you were handed.

**Pasted text.** Return the rewrite, then a short list of what you changed and
anything still weak. Show the work.

**File path.** Rewrite prose in place. Leave code blocks, frontmatter, data, config,
link targets, and command output exactly as they are. Report a two-line summary.

**Embedded.** Another skill called you for a title, caption, commit body, or PR
description. Return the finished text alone. No commentary, no preamble.

## The pass

1. Read the source. Mark every signal from §A through §F.
2. Rewrite whole. Do not patch flagged phrases one at a time, that leaves a draft
   that reads like it was edited by a machine to sound less like a machine.
3. Read it back and ask two questions:
   - What still sounds generated?
   - Did anything get added or lost?
4. Fix what those two questions surface. Then check §C1 for stray dashes.

Step 2 is where most passes fail. A phrase-level patch keeps the underlying rhythm.
Rebuild the paragraph.

## §A Claims and sourcing

**A1. Manufactured significance.** Ordinary facts described as turning points,
legacies, or evidence of a broader shift. Watch: marks a pivotal moment, stands as
a testament, underscores the importance of, reflects a broader, cements its place,
a defining chapter.

- Before: The agency opened a second office in 2023, marking a pivotal moment in its
  evolution and underscoring its commitment to regional growth.
- After: The agency opened a second office in 2023.

**A2. Unnamed authorities.** A claim handed to experts, observers, critics, or
studies that are never named. Either name the real source or cut the claim. Never
invent one.

- Before: Industry experts agree that short-form video drives the strongest returns.
- After: Cut it, or cite the actual report and its number.

**A3. Credential padding.** Logo lists and follower counts standing in for a reason
to care. Keep a citation that carries information. Drop the ones that only signal.

**A4. Filled gaps.** The model could not find something, said so, then guessed
anyway. Watch: while details are limited, based on available information, likely
began, it is believed that, appears to have. State what the source does not show,
or cut the sentence. A guess presented as a fact is the worst failure mode here.

**A5. Stacked hedges.** Qualifiers piled until nothing is claimed. "It could
potentially be argued that this may have some effect" is one uncertain claim wearing
four hedges. Keep the hedge the evidence supports. Delete the rest.

## §B Verbs and sentence shape

**B1. Verb avoidance.** *Serves as, stands as, represents, features, boasts,
offers* where *is* and *has* would do.

- Before: The dashboard serves as a central hub and features four distinct views.
- After: The dashboard is the central hub. It has four views.

**B2. Brochure vocabulary.** vibrant, robust, seamless, nestled, breathtaking,
renowned, must-have, stunning, groundbreaking, rich (figurative), in the heart of.
Also the reliable tells: delve, crucial, pivotal, comprehensive, testament,
tapestry, landscape (abstract), showcase, underscore, foster, garner, interplay,
intricate, quietly, enduring.

**B3. The -ing tail.** A participle clause bolted on to make a plain fact sound
analyzed. Watch: highlighting, reflecting, symbolizing, ensuring, showcasing,
fostering, contributing to, underscoring.

- Before: The page loads in under a second, reflecting the team's commitment to
  performance.
- After: The page loads in under a second.

**B4. Not X but Y.** "It's not just a tool, it's a workflow." Also the clipped
negative tail: ", no guesswork", ", no setup required". State the thing.

- Before: The importer maps columns automatically, no manual mapping.
- After: The importer maps the columns for you.

**B5. Forced triads.** Three items because three sounds finished, not because there
are three. Use the number the meaning needs. Two is allowed. So is four.

**B6. Fake ranges.** "From X to Y" where X and Y are not ends of any spectrum.
"Everything from onboarding to dark mode" is a list wearing a range. Just list them.

**B7. Buried actors.** Passive voice or a dropped subject hiding who does what.
"The results are preserved automatically" leaves out what preserves them.

**B8. Synonym cycling and stuck openings.** Renaming one subject every sentence
(the founder, the entrepreneur, the executive) or starting five sentences with the
same subject. Fix the pattern, not the word. Merging two sentences usually solves it.
Deliberate anaphora for rhythm is not this.

## §C Format

From ronnnerr/norstack, MIT.

Signals

Stars
1
Forks
0
Open issues
0
Last commit (this folder)
Aug 28, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

Version
1.0.0
License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from ronnnerr

Cite this page

humanizer skill. Agent Mart. https://agentmart.dev/skills/ronnnerr-humanizer (last updated 2026-10-24)