Ships as reference material with no one-line installer — clone it and copy the folder into .claude/skills/. The rules synthesise published methodologies; they are not measured against your readers.
why it matters
Where an auditor scores what you have, this decides what to write. Diátaxis’
four modes expand into the fourteen types a developer product actually ships —
quickstart, migration guide, troubleshooting, SDK reference, runbook. The
pluggable style guides are the pragmatic part: pick Google or Stripe and
inherit a settled answer on voice, person and tense instead of re-running that
argument in every review.
Structural checks cannot tell you whether the quickstart runs. Pair the score with a real time-to-first-call test on a clean machine.
why it matters
Docs are the highest-leverage surface you own and nobody scores their own. A
thirty-three-check pass with a number at the end is worth having because it
turns “our docs are fine” into a claim someone can disagree with before a
launch, and because its AI-discoverability half is work most docs teams have
not been asked to do yet.
One artifact only — it grades the README, not the docs site behind it, and the voice is yours to supply via a style profile.
disclosure
Published by this site's author.
why it matters
For most developer products the README is the landing page: where a link
from Hacker News lands, and what a model retrieves when someone asks what your
thing does. It is the only skill here that treats it as a scored artifact,
every rule traced to a source and review returning concrete replacement text
for the worst offender. Reach for it before a launch, and again the moment an
install command changes.
Anthropic's own four-phase guide to building an MCP server — research, implement, review, then generate ten evaluations that test whether a model can actually use it.
An engineering skill, engineering-owned in most teams. Bring it to whoever ships the server rather than running it alone.
why it matters
For a growing share of your users the tool description is the documentation:
they never read the reference, because an agent chose your tool on one
sentence and a JSON schema. That makes naming, annotations and error copy a
first-touch surface nobody else in the company will own. The phase worth
stealing is the last: ten verifiable evaluations that measure whether a model
can complete real tasks against your server.
Scores SDK docs across the six sections integration actually fails in — installation, quick start, error handling, troubleshooting, examples, best practices — as an HTML report.
It audits the docs for the SDK, not the SDK. A clean report on a hostile API is still a hostile API.
why it matters
Activation rarely dies at install; it dies in the hour after, when the first
call returns an error nobody documented. The six sections this grades map onto
that hour, and the two teams usually miss — error handling and troubleshooting
— are the two nobody is measured on. Run it on your own SDK docs and on the
competitor a prospect is comparing you against.
The knowledge layer for generative-engine optimisation — content patterns, structured data, crawler config, llms.txt, E-E-A-T and the monitoring to see if any of it moved.
install
npx skills add luka2chat/geo-skills
caveat
No citations behind the rules — it descends from a curated awesome-list. A well-organised checklist, not evidence.
why it matters
Audit tools tell you what is missing; this changes how you brief the work in
the first place. Twelve rules across content, technical implementation,
authority and platforms, each with templates and config you can lift. It is
here because the rules are organised the way a marketer thinks rather than the
way a crawler works.
Twelve skills that research, plan and then patch — AEO/GEO auditing with framework-specific fixes, shipped with worked plans for a Next.js blog and a WordPress site.
Manual clone-and-symlink per skill, and a marketing agent editing your app needs the same review as any other contributor.
why it matters
The third GEO entry here, and the one that goes past reporting: improve-aeo- geo proposes framework-specific changes to the site itself. Pick between the
three by what you are short of — the knowledge, the number, or the diff. The
marketing engineer it assumes, comfortable opening a pull request against the
marketing site, is the role this whole shelf quietly requires.
Presence, not persuasion — every file can pass while the content behind them earns no citation.
why it matters
The cheapest check on this page: minutes, no credentials, and it answers a
question most devtool teams have not formally asked — whether the systems now
fielding “what should I use for X” can reach and parse you at all. Treat a
pass as a floor rather than a strategy: being legible is not being chosen.
More than thirty skills covering audience context, content, distribution, developer experience, growth and retention — every one reads a shared developer-audience file first.
install
npx add-skill jonathimer/devmarketing-skills
caveat
Prescriptive playbooks, not sourced research — the outputs are drafts, and the claims inside them are yours to check before they ship.
why it matters
The closest thing to this guide in installable form, grouped the way the work
divides — audience, content, distribution, developer experience, growth,
lifecycle. The move worth stealing is developer-audience-context: state your
persona, their technical level and their pain points once, and every other
skill reads it first, which is what stops the output drifting into generic
SaaS voice.
A repeatable sweep of developer conversation across Reddit, Hacker News, Discord and issue trackers, turned into notes a DevRel team can act on.
install
npx add-skill jonathimer/devmarketing-skills
caveat
Public conversation only — the sharpest signal about your own product still arrives in the support queue and the issue tracker.
why it matters
DevRel’s real input is what developers say when you are not in the room, and
most teams do it badly because the work is unglamorous, cross-platform and
never finished. Handing the sweep to an agent makes the cadence survivable;
its siblings in the repo cover what to do once you have heard something.
Scaffolds and validates an Agent Skill — the packaging format that puts your conventions, gotchas and idiomatic examples inside the agent your users already work in.
Ship knowledge that is genuinely yours. A skill that reads like a brochure gets uninstalled faster than a bad README gets closed.
why it matters
A skill is a content surface that did not exist two years ago: your migration
gotchas, your idiomatic error handling, installed into the tool a developer is
already in and invoked without them ever visiting your site. This builds and
validates one, and the discipline it enforces is the discipline good docs need
anyway — a description tight enough to trigger on the right task, and a body
small enough to stay read.
Reads git history, groups changes into added, improved, fixed and breaking, drops internal noise, and returns release notes a human can edit in ten minutes.
Commits are the wrong source for why a change matters. It drafts the what; the why, the migration note and the breaking-change warning are yours.
why it matters
The changelog is read more often than the blog and owned by nobody, because
turning forty commits into user-facing language loses to whatever shipped that
week. An agent draft restores the cadence — but keep the human edit: the
categories come from the commits, “change your webhook handler before August”
never does.
The general-marketing shelf — 71 skills across CRO, copywriting, SEO, pricing and lifecycle — wired so each one reads your product-marketing context file before it writes.
install
npx skills add coreyhaines31/marketingskills
caveat
Written for marketers at large, so it will hand you funnel advice a developer audience punishes. Take the structure; keep the technical judgment yours.
why it matters
The most-used shelf of the lot — 51k stars as of September 2026 — and it earns
its place here for a structural reason: every skill reads product-marketing
first, one file holding your product, audience and positioning, so the claim
is stated once and inherited rather than re-improvised on every page. The
skills are generic B2B by default; that context file is where you make them
survive contact with an engineer.