Humanizer or Stop Slop Claude skill, which is better?

For editing text that already exists, Humanizer is better as of October 2026: it checks 26 patterns based on Wikipedia's 'Signs of AI writing', matches your voice from a sample, refuses to invent facts, installs as a Claude Code plugin and is actively maintained (v3.1.0, updated September 28, 2026, about 54,000 GitHub stars). Stop Slop is better as a strict, lightweight rulebook to put in a system prompt so the model writes cleaner in the first place: it bans specific phrases and structures and scores drafts on five dimensions. Many writers use Stop Slop's rules while drafting and Humanizer for the final edit.

Do Humanizer or Stop Slop help bypass AI detectors?

No, and neither claims to. Humanizer's README says getting past AI detectors is not a goal and that detectors still flag most of its output. Both skills aim at prose that reads better to people. If you are evaluating detection itself, test detectors on your own samples.

How do I install Humanizer in Claude Code?

Run /plugin marketplace add blader/humanizer and then /plugin install humanizer@humanizer (Claude Code 2.1.142 or newer). On older versions or other agents, use npx skills add blader/humanizer. Stop Slop is added by putting its folder in your skills directory, uploading SKILL.md to a Claude Project, or pasting its rules into a system prompt.

What is the difference between Humanizer and Stop Slop?

Humanizer is an editing procedure: mark tells strongest-first, draft a rewrite, check it did not add or drop facts, then finalise, with voice matching from a sample. Stop Slop is a set of hard rules plus reference lists of banned phrases and structures, with a 1-10 score on directness, rhythm, trust, authenticity and density and a revise threshold below 35 out of 50. Stop Slop is stricter (it bans all adverbs, em dashes and passive voice); Humanizer weighs patterns and lets a writing sample override its dash rule.

Quick Answer

Humanizer vs Stop Slop: Which Claude Skill Is Better?

Published: October 6, 2026

The short answer

Use Humanizer to edit, Stop Slop to prevent. Humanizer is the better tool for turning an AI-written draft into prose that sounds like you, because it works through a check-and-rewrite procedure, matches a voice sample and guards against invented facts. Stop Slop is the better tool to load into a system prompt so drafts come out cleaner, because it is a short, strict rulebook. They don’t conflict; plenty of people run both.

Side by side (October 2026)

	Humanizer (blader/humanizer)	Stop Slop (hardikpandya/stop-slop)
What it is	Agent skill that rewrites text to remove AI tells	Skill file of rules for removing AI tells from prose
Basis	Wikipedia’s “Signs of AI writing”; 26 patterns, strongest first	Author’s banned phrases, structures and sentence rules
Approach	Mark tells → draft rewrite → check facts and tells → final	Hard rules + 5-dimension score (revise below 35/50)
Voice matching	Yes, from a 2–3 paragraph sample	No
Fact safety	Explicit: never add names, numbers, dates, quotes or citations	Not addressed
Strictness	Weighted; some patterns “weak alone”	Strict: no em dashes, no adverbs, no passive voice, no “Wh-” openers
Install	Claude Code plugin; `npx skills add` for Codex, Gemini CLI, Copilot, Windsurf	Copy folder as a skill, upload to Projects, or paste into a system prompt
Licence	MIT	MIT
Activity	v3.1.0; last commit Sep 28, 2026; ~54,000 stars	Last commit Mar 17, 2026; ~17,800 stars
Evidence	Blind test: judges preferred its rewrite 16 of 16 times (issue #229)	Self-scoring rubric

Humanizer: the better editor

Humanizer’s skill file explains AI prose as “the default choice”: a model picks what fits the widest range of readers, so you get staged contrasts (“it’s not just X, it’s Y”), one-line closers, forced triads, inflated claims and leftover chat wrappers. It ranks 26 patterns; the top five justify an edit on one sighting, weaker ones only in combination. The procedure matters most: after drafting, it re-checks whether the rewrite added or dropped any fact, name, number, date, quote or ranking, and asks rather than inventing a missing detail. With a writing sample it follows your rhythm and punctuation, including dashes if you use them.

Install in Claude Code with /plugin marketplace add blader/humanizer and /plugin install humanizer@humanizer (Claude Code 2.1.142+), then call /humanizer or say “humanize the prose in docs/launch-post.md”.

Stop Slop: the better guardrail

Stop Slop is short on purpose: SKILL.md plus three reference files — phrases to remove, structural patterns to avoid, and before/after examples. It bans throat-clearing openers, emphasis crutches, business jargon, all adverbs, binary contrasts, dramatic fragmentation, rhetorical setups and passive voice, and it scores a draft 1–10 on directness, rhythm, trust, authenticity and density, revising anything under 35 of 50. Because it is plain Markdown with no code, it drops into Claude Projects, custom instructions or an API system prompt.

The trade-off is bluntness. “No adverbs” and “no em dashes” remove tells but also remove legitimate style, and there is no rule about preserving facts. Updates stopped in March 2026.

Which to use

You have an AI draft and want it to sound like you: Humanizer with a voice sample.
You want every draft cleaner from the start: Stop Slop in the system prompt or project knowledge.
Technical or factual content (docs, reports, posts with numbers): Humanizer, for its fact-preservation check.
Scientific writing: look at skill-deslop, a smaller skill built partly from Stop Slop for academic prose.
Pipelines: Stop Slop rules at generation, Humanizer as the final pass, then a human read.

Neither is a detector-evasion tool. Judge results by reading them, not by an AI-detector score.

Last verified: October 6, 2026. Star counts and commit dates from GitHub on that date.

The short answer

Side by side (October 2026)

Humanizer: the better editor

Stop Slop: the better guardrail

Which to use

Sources