← Skill Arena

Docs & Writing skills

Skills that draft long-form docs, blog posts, specs and copy.

We gave six writing skills the same three briefs, all written by the same base model. Every piece was judged blind on the text by two independent judges (the previews here are that same text rendered as a clean article). Because the model is held constant, the differences reflect the skill.

Model: Claude Sonnet 56 competitors3 tasks2 blind judges

Leaderboard

Rank 1
copywriting

coreyhaines31 · Almost anything customer-facing, its clarity-first discipline won all three tasks and the arena overall.

9.0
Rank 2
beautiful-prose

SHADOWPR0 · Narrative and persuasive writing that must read sharp and human, strong on the launch.

8.3
Rank 3
doc-coauthoring

Anthropic · Technical explainers and structured docs where accuracy and structure matter.

8.0
4
writing-prds

refoundai · Product specs and PRDs with crisp problem, scope and metrics.

8.0
5
internal-comms

Anthropic · Team announcements, updates and memos with a calm, credible voice.

7.3
6
blog-write

AgriciDaniel · Full SEO blog articles where metadata, FAQ and internal links matter.

4.2

Overall = mean of the per-task overall scores from 2 blind judges.

Every result, side by side

Every skill built the same brief on the same model. Tap any result to open the live page.

Technical blog post

Winner: copywriting

An ~800-word technical blog post explaining how vector databases work for a developer audience, with a clear intro, concept sections, a concrete example and a takeaway.

9.0Open live ↗
Rank 1 for this taskcopywriting

Punchy, developer-native voice with a runnable end-to-end example including realistic output scores; hits every brief beat cleanly.

Open live demo ↗
8.0Open live ↗
Rank 2 for this taskbeautiful-prose

Strong distinctive prose voice; the concrete example is only a query snippet without ingest, and drifts past the word target.

Open live demo ↗
8.0Open live ↗
Rank 3 for this taskdoc-coauthoring

Most technically precise (ef_construction/ef_search, IVF+PQ, recall@10) but denser and slightly over-length; example less illustrative than A.

Open live demo ↗
8.0Open live ↗
writing-prds

Solid and accurate with useful comparison table, but adds hybrid/filtering sections that push well past ~800 words and dilute focus.

Open live demo ↗
8.0Open live ↗
internal-comms

Clean staged structure and tight runnable example; solid but slightly more generic and safe than the top pieces.

Open live demo ↗
5.5Open live ↗
blog-write

SEO-article format (frontmatter, FAQ, internal-link placeholders, citation capsules) is off-brief for a technical blog post and pads well past 800 words.

Open live demo ↗

Product launch announcement

Winner: copywriting

A ~450-word launch announcement for an AI note-taking app, a headline, the news, three features with benefits, a customer quote and a clear CTA.

9.0Open live ↗
Rank 1 for this taskcopywriting

Best opening hook; 'What you get' framing plus specific numbers and a metric-driven quote make it the most persuasive and tightest fit for the brief.

Open live demo ↗
9.0Open live ↗
Rank 2 for this taskbeautiful-prose

Punchy, tight, on-brand launch voice; strong features and quote, though quote is short and no explicit benefit labels—still the cleanest fit for the brief.

Open live demo ↗
8.0Open live ↗
Rank 3 for this taskdoc-coauthoring

Well-structured with clear feature/benefit pairs and specific quote; privacy feature is a nice differentiator but runs slightly long past 450 words.

Open live demo ↗
8.0Open live ↗
writing-prds

Polished and complete with all brief elements; slightly long and headline-heavy with extra 'Success Looks Like' section, but tight prose and strong benefits.

Open live demo ↗
7.0Open live ↗
internal-comms

Solid and credible with concrete stats, but features lack explicit benefit callouts and no clear standalone CTA button; reads more like an update memo.

Open live demo ↗
3.0Open live ↗
blog-write

A ~1500-word SEO blog with frontmatter, FAQ, tables, sources and internal-link placeholders—explicitly off-brief format despite competent content.

Open live demo ↗

One-page PRD

Winner: copywriting

A one-page PRD for a Shared Workspaces feature, problem, goals, non-goals, user stories, scope, success metrics and risks.

9.0Open live ↗
Rank 1 for this taskcopywriting

Crisp, skimmable, punchy human user stories and concrete metrics; hits every PRD section cleanly.

Open live demo ↗
8.0Open live ↗
Rank 2 for this taskbeautiful-prose

Vivid problem framing and good scope; narrative user stories lean slightly literary but stay on-brief.

Open live demo ↗
8.0Open live ↗
Rank 3 for this taskdoc-coauthoring

Strong tables (stories, risks with likelihood/impact), specific engineering targets like P99 latency; dense but tight.

Open live demo ↗
8.0Open live ↗
writing-prds

Thorough with timeline and open questions, but slightly long for a one-pager; solid and well-organized.

Open live demo ↗
7.0Open live ↗
internal-comms

Clean and skimmable but thinner: fewer user stories/risks and lighter metrics than A/C; solid but modest.

Open live demo ↗
4.0Open live ↗
blog-write

Off-brief: SEO frontmatter, Key Takeaways, citations, FAQ, backlinks — a blog masquerading as a PRD, not skimmable one-pager.

Open live demo ↗

How we scored docs & writing

Two independent blind judges scored every output 1 to 10 on each of these axes, then we averaged across judges and tasks. These axes are specific to docs & writing, while other arenas use their own rubric.

Clarity
Easy to follow, no fluff or filler.
Structure & flow
Logical flow and skimmable organization.
Substance
Accurate, specific and non-generic content.
Voice & fit
Tone and format fit for the brief and audience.
Completeness
Covers everything the brief asked for.

Limitations. This arena is a single run on Claude Sonnet 5 with 3 tasks, scored by AI judges rather than crowd votes. Treat the results as directional evidence, not a definitive ranking; re-test on your own workload before committing.

How each skill performed, and who it’s for

9.0
rank #1

Marketing copywriting skill: hooks, benefit-led structure and conversion-focused, jargon-free copy.

Clarity
9.0
Structure & flow
9.0
Substance
8.8
Voice & fit
9.0
Completeness
9.0

Best for: Almost anything customer-facing, its clarity-first discipline won all three tasks and the arena overall.

Watch out: Tuned for a marketing register; may feel too punchy for a formal spec.

beautiful-prose

SHADOWPR0

8.3
rank #2

A hard-edged prose style contract for forceful, timeless English without modern AI tics.

Clarity
8.3
Structure & flow
7.7
Substance
8.0
Voice & fit
8.8
Completeness
7.7

Best for: Narrative and persuasive writing that must read sharp and human, strong on the launch.

Watch out: A style layer, not a structure engine, pair it with an outline for complex docs.

doc-coauthoring

Anthropic

8.0
rank #3

Anthropic's structured doc co-authoring workflow: plans, then drafts documentation and specs.

Clarity
8.8
Structure & flow
8.3
Substance
8.8
Voice & fit
7.8
Completeness
8.7

Best for: Technical explainers and structured docs where accuracy and structure matter.

Watch out: Process-oriented; heavier than needed for short punchy copy.

writing-prds

refoundai

8.0
rank #4

Turns abstract ideas into actionable PRDs and specs that align engineering and design.

Clarity
8.3
Structure & flow
8.0
Substance
8.3
Voice & fit
7.8
Completeness
8.8

Best for: Product specs and PRDs with crisp problem, scope and metrics.

Watch out: Spec-first voice can read dry for persuasive or narrative pieces.

internal-comms

Anthropic

7.3
rank #5

Anthropic's internal-comms skill: clear announcements, updates and memos for a team audience.

Clarity
8.3
Structure & flow
8.0
Substance
7.7
Voice & fit
7.8
Completeness
7.5

Best for: Team announcements, updates and memos with a calm, credible voice.

Watch out: Reads memo-like for punchy external launches or deep technical posts.

4.2
rank #6

SEO-oriented blog writing: frontmatter, on-page structure, FAQ and internal-link scaffolding.

Clarity
6.5
Structure & flow
6.0
Substance
6.5
Voice & fit
2.8
Completeness
6.7

Best for: Full SEO blog articles where metadata, FAQ and internal links matter.

Watch out: Forces an SEO-blog template even when it's off-brief, badly wrong for launches and PRDs (its lowest scores here).

The tasks, head-to-head

How each skill scored on the three briefs, and how quality trades off against token cost.

Scores by task
SkillTechnical blog postProduct launch announcementOne-page PRDAvg
copywriting9.09.09.09.0
beautiful-prose8.09.08.08.3
doc-coauthoring8.08.08.08.0
writing-prds8.08.08.08.0
internal-comms8.07.07.07.3
blog-write5.53.04.04.2

Cell = the skill’s score on that task (darker = higher). A ring marks the task winner.

Cost vs quality— top-left = cheap and high-scoring
3.06.510.04K10K26KContext tokens per run (log scale, cheaper ←)Quality (overall, better ↑)copywritingbeautiful-prosedoc-coauthoringwriting-prdsinternal-commsblog-write

Token cost & efficiency

Every skill ran on the same base model, so token use reflects the skill itself: how much it writes (generated) and how large its SKILL.md is (context processed). Summed across the 3 tasks; lower is cheaper.

GeneratedContext (total processed)
internal-commsquality #5
Generated
3K
Context
4K
beautiful-prosequality #2
Generated
3K
Context
7K
copywritingquality #1
Generated
3K
Context
9K
writing-prdsquality #4
Generated
4K
Context
9K
doc-coauthoringquality #3
Generated
4K
Context
15K
blog-writequality #6
Generated
8K
Context
26K

Context includes the skill’s SKILL.md re-read into the model each turn, so a larger skill file drives up cost, while the no-skill baseline is cheapest.

What we learned

  • copywriting won all three briefs and the arena overall (9.0), a clarity-first, benefit-led marketing discipline generalizes well beyond marketing copy.
  • beautiful-prose was the runner-up (8.33) and strongest on the launch, a good style layer sharpens almost any piece.
  • doc-coauthoring and writing-prds are dependable for structured work (explainers, specs) but read a little flatter on persuasive briefs.
  • internal-comms is solid but memo-like, fine for updates, less punchy for external launches or deep technical posts.
  • Format fit matters most at the bottom: blog-write forced an SEO-blog template (frontmatter, FAQ, internal-link stubs) onto every brief, which tanked it on the launch and PRD (3.0 and 4.0).

Frequently asked questions

What is the best AI skill for writing?+

In our blind test copywriting (coreyhaines31) scored highest overall (9.0/10), winning all three briefs, a blog post, a launch announcement and a PRD, on the strength of clear, benefit-led, jargon-free prose. beautiful-prose (SHADOWPR0) was a close runner-up (8.33).

Which AI skill is best for blog posts, marketing copy, or PRDs?+

In our tests copywriting won all three (technical blog, marketing launch, and one-page PRD) by writing clearly and on-format, with beautiful-prose especially strong on the launch and doc-coauthoring and writing-prds solid on the structured pieces. Match the skill to the format, but a clarity-first copy skill travels surprisingly well.

Do writing skills actually improve the output?+

Yes, mostly through clarity, voice and format fit. A strong copy or prose skill sharpens almost any piece, while format-locked skills help only for their format. The clearest lesson is negative: an SEO-blog skill hurt badly when the brief was a launch or a PRD because it forced the wrong structure.

How is the Docs & Writing arena scored, and is it fair?+

Every skill writes the same three briefs with the same base model, so results isolate the skill, not the model. Each piece is anonymized and scored 1 to 10 by two independent blind judges on clarity, structure, substance, voice/fit and completeness, then averaged across judges and tasks.

Where can I find and install these writing skills?+

All are openly available: copywriting (coreyhaines31), beautiful-prose (SHADOWPR0), doc-coauthoring and internal-comms (Anthropic), writing-prds (refoundai / Lenny's), and blog-write (AgriciDaniel). On Happycapy you can install any of them and run them in your browser with no setup.

Build with these skills, no setup required

Every skill in this docs & writing arena installs on Happycapy in one click and runs in your browser on the same model. Start free, or explore 2M+ skills in the Skill Store.