Best Claude Skills for Writing (2026)

Most "best skills" lists are copied README summaries. This one isn't: we installed all 168 writing skills in our catalog, ran each against a baseline and scored four criteria. 93 passed cleanly, average score 9.0/10. The one to start with: Humanizer.

Ranked by SkillProof Score · how we score

  1. 1

    Humanizer

    Strips 33 documented AI-writing tells from text via a draft-audit-final rewrite loop.

    From our test: Ran a sample AI-sounding product-launch paragraph through the draft→audit→final loop: the draft still had a stray em dash and a generic closer, and the mandatory second pass caught and fixed both — a…

    Tested · Works ★ 9.6/10 by blader

  2. 2

    Blog Outline Generator

    SERP-informed H2/H3 outline generator; skeleton only, not a full content brief.

    From our test: Live-tested on 'best project management software for remote teams': real SERP research surfaced the market's dominant 'best-for-X' comparison-table-plus-FAQ pattern (confirmed by fetching a…

    Tested · Works ★ 9.6/10 by AgriciDaniel

  3. 3

    Academic Paper Reviewer

    Simulates rigorous academic peer review across originality, methodology, results, and writing.

    From our test: Turned a freeform review paragraph into a structured 4-dimension writeup with Major/Minor tags and a priority table - markedly more actionable than the unguided version.

    Tested · Works ★ 9.6/10 by zebbern

  4. 4

    PR Writing Review

    Extracts before/after writing fixes and style lessons from GitHub PR review comments.

    From our test: Against a real merged PR, its extraction script pulled 21 suggestion blocks and correctly tracked the target file through 3 renames; a raw API call returned the same comments flat and unstructured.

    Tested · Works ★ 9.6/10 by evalstate

  5. 5

    Chapter Writing

    Outline-first chapter drafting that reads your whole story bible and updates continuity after writing.

    From our test: Ran a real side-by-side on a scratch story project: the no-skill baseline chapter opened with 'Rain hammered the tin roofs...' — the exact cliche the skill's own writing-guidelines.md tells you to…

    Tested · Works ★ 9.6/10 by danjdewhurst

  6. 6

    Article Writing

    Bans 'rapidly evolving landscape' and 'game-changer', forces proof-first sections in long-form drafts

    From our test: Wrote the same Python-venv guide intro twice: the baseline opened with 'in today's rapidly evolving software landscape' and closed on 'let's dive in!' — exactly the two patterns the skill's own…

    Tested · Works ★ 9.6/10 by mturac

  7. 7

    Doc Co-Authoring

    Three-stage doc-writing workflow — context dump, section-by-section drafting, then a fresh-Claude reader test — instead of drafting blind.

    From our test: Asked for help with a monorepo decision doc: unguided Claude jumps straight into drafting a doc filled with invented specifics (team size, CI system) it doesn't actually know; following the skill it…

    Tested · Works ★ 9.6/10 by ThinkInAIXYZ

  8. 8

    AI Video Prompt Director

    Forces a mandatory dramaturgy + universal-rules + model-file reading order before writing any AI video prompt

    From our test: Fed it a bare 'detective in a rainy alley' prompt request: the mandatory dramaturgy + universal-rules reading order forced a motivated camera move, a specific physical micro-action, and Seedance's…

    Tested · Works ★ 9.6/10 by smixs

  9. 9

    Chinese Reference Formatter

    Verified-metadata GB/T 7714 bibliography and BibTeX from a bare title.

    From our test: Ran the bundled format_reference.py on a real paper's metadata and got a correctly GB/T 7714-formatted entry plus BibTeX in one shot — author truncation to '等/et al', no-period initials, and the…

    Tested · Works ★ 9.6/10 by Zechang-Xiong

  10. 10

    OneCite

    Verifies references against live academic APIs and emits clean, complete BibTeX.

    From our test: Ran two real DOIs through it live: got back correct, fully-formed BibTeX (author, journal, volume, pages, abstract, DOI, URL) pulled from actual metadata APIs, not memorized text -- but the…

    Works with setup ★ 9.6/10 by HzaCode

How this ranking works

Each skill gets a fresh install in Claude Code, three trigger phrasings, and the same task run with and without it. Scores: installation (5), trigger reliability (5), output vs baseline (10), documentation honesty (5) — normalized to a 10-point SkillProof Score. Full rubric on the methodology page. Queue-verdict skills sit at the bottom until testing completes.

New here? Start with how to install Claude skills — or grab the whole setup in one command with a curated pack ($10).

On a different AI tool? These tested instructions adapt: see our conversion guides for Cursor rules, Copilot instructions, Windsurf and Custom GPTs.