High 5 Claude Abilities for Writing (Ranked by GitHub Stars)

Search “finest Claude Abilities for writing” and also you get lists padded with expertise that write commit messages and inner standing reviews. Helpful issues. Not writing.

This record solely contains repositories that exist for writing. Each entry is a repository whose total cause for being is writing or modifying, which suggests its star depend measures the factor you care about. Listed below are a few of the finest writing agent expertise (ranked by their reputation) that assist distinguish your writing from a mere: AI-slop!

Notice: All figures pulled from the GitHub on fifth August 2026.

1. humanizer | ★ 33,700

blader/humanizer · 1,005 forks

Thirty-three numbered patterns, every with a earlier than and after instance, all drawn from Wikipedia’s Indicators of AI Writing information. The early variations went after apparent vocabulary, however later releases moved to rhythm: subjectless fragments, manufactured punchlines, the chatty rhetorical opener that solutions its personal query, and em dashes reduce outright slightly than merely rationed.

Two issues it does that stop-slop doesn’t:

Voice matching: Paste two or three paragraphs of your individual writing first. It reads your sentence rhythm, phrase decisions, and quirks, then rewrites towards these as a substitute of towards generic clear prose. Most anti-slop expertise flatten every little thing to the identical impartial register. This one goals at yours.

A second move: After rewriting, it runs an “clearly AI generated” audit over its personal output and rewrites once more, on the speculation that the primary move leaves tells.

Finest for: Every little thing. Default Claude responses are on the verge of AI-slop themselves. Including this expertise, helps giving them a humane contact.

2. stop-slop | ★ 15,199

hardikpandya/stop-slop · 1,082 forks

Essentially the most-starred writing ability on GitHub, and it’s one SKILL.md plus three reference recordsdata. It really works as a result of it refuses to inform the mannequin to “sound extra human.” It ships lists as a substitute:

phrases.md bans throat-clearing openers, emphasis crutches, enterprise jargon, each adverb, obscure declaratives, meta-commentary

constructions.md bans binary contrasts, unfavorable listings, dramatic fragmentation, rhetorical setups, false company, passive voice

Sentence guidelines bans em dashes, staccato fragments, and lazy extremes like “each” and “by no means”

Then it scores the end result 1 to 10 on directness, rhythm, belief, authenticity, and density. Under 35 out of fifty, revise. That’s a verify the mannequin runs towards itself, which is what separates a ability from a suggestion.

Finest for: Last move on something drafted with AI assist.

3. AI-Analysis-SKILLs | ★ 11,400

Orchestra-Analysis/AI-Analysis-SKILLs · 833 forks

Ninety-eight expertise throughout 23 classes, constructed to let an agent run the complete AI analysis lifecycle from literature survey to submitted paper. Most of it’s engineering: Megatron, vLLM, TRL, Flash Consideration, quantization. Not writing.

However three of its classes are, and so they’re stronger than most standalone writing repos:

Educational Plotting generates publication-quality figures with venue-specific styling, protecting each structure diagrams and matplotlib knowledge charts.

Ideation has two expertise for the blank-page downside. Analysis Brainstorming runs ten complementary lenses over a subject. Inventive Considering applies bisociation, structure-mapping and constraint manipulation.

ARA Rigor Reviewer scores a draft throughout six dimensions of analysis rigor, together with falsifiability, scope calibration, argument coherence and methodological rigor, then points a severity-ranked report ending in an accept-to-reject advice.

4. oh-story-claudecode | ★ 5,080

worldwonderer/oh-story-claudecode · 687 forks

13 instances the celebs of my fiction entry, and a greater concept of what fiction writing with an agent seems like. Constructed for Chinese language net fiction, however the English README is full and the tactic transfers.

The pipeline is development scouting, deconstruction, writing, slop cleanup, cowl artwork. The center step is the one no person else does. Level it at a novel that labored and it produces a structured teardown:

Story Teardown Report scores the story core throughout 5 dimensions, six payoff sides, cognitive reversal, and 9 layers of resonance.

Plot Beats extracts each plot level with an emotion marker from −9 to +9, supplying you with the emotional curve as knowledge.

Craft Methods catalogues eleven strategies: POV dealing with, dialogue, data gaps, object hooks

These outputs change into a reusable library. The writing expertise learn from it, so that you draft towards the construction of one thing that already labored slightly than towards the mannequin’s defaults.

5. claude-blog | ★ 1,579

AgriciDaniel/claude-blog · 274 forks

Thirty sub-skills and 5 brokers protecting the entire weblog lifecycle: technique, briefs, outlines, drafting, rewriting, schema, website audits, matter clusters, content material decay detection.

The half price stealing even for those who by no means set up it’s the supply contract. Each draft is scored towards a 100-point rubric throughout 5 gates, and something under 90 is blocked and re-run as much as 3 times earlier than it reaches you. The creator’s framing: the person is rarely the primary reviewer.

That’s the verification sample utilized to prose. Most writing expertise cease when the output seems completed, which makes you the standard gate.

Finest for: Publishing at quantity, the place you possibly can’t personally evaluate each draft.

Putting in them in Claude Code

Set up any of those by cloning into your expertise listing:

git clone https://github.com// ~/.claude/expertise/

Last Ideas

Choose by what you write:

Everybody ought to have humanizer regardless. It’s the one entry that improves output it doesn’t matter what the output is (overlook the em-dashes and different AI markers).

Learn extra: The way to Join MCP Servers with Claude (Claude Desktop and Claude Code)

Incessantly Requested Questions