AI Marketing Skills — research, SEO/GEO, analytics & content

Name: marketing-skills
Author: LuCheremisina

Published v2.0.7: 34 authored CHE_ workflows + 14 licensed engineering skills. Release · Migration · MCP Panel connections.

Install the complete plugin or one skill: Codex, Claude Code and Cursor commands · Claude / Cowork and ChatGPT.

Platform guide: installation options and compatibility limitations.
Research a market, audit a website, analyse Yandex campaigns, find search demand, or turn evidence into an expert article with reusable AI workflows.

34 original marketing workflows + 14 optional engineering skills. Open Agent Skills packages for Codex, Claude Code and Cursor; ChatGPT plugin packages are also available. Each workflow lists the inputs and connections it needs. Platform status.

Install the complete plugin · One skill · How to use · Start with 5 marketing skills · Browse all 48 · Download the latest release · Русская версия

Quick install

Choose one route for your system. The complete plugin bundles all 34 original marketing skills. The npx skills CLI installs the same workflows as individual skill folders, without registering a plugin. The 14 third-party engineering skills are an optional separate add-on.

Complete plugin

Claude Code — enter these commands in the Claude Code chat, in order:

/plugin marketplace add LuCheremisina/marketing-skills
/plugin install marketing-skills@cheremisina-marketing

Codex CLI — run in your terminal:

codex plugin marketplace add LuCheremisina/marketing-skills
codex plugin add marketing-skills@cheremisina-marketing

In Codex with a plugin browser, open Plugins → Add plugin marketplace, enter LuCheremisina/marketing-skills, then install Marketing Skill Cheremisina. If that control or command is unavailable in your version, use the Skills route below.

Claude / Cowork — open Customize → Plugins → Add → Add marketplace → Add from a repository, enter https://github.com/LuCheremisina/marketing-skills, then select and install the plugin. Alternatively, download the ZIP with all 34 skills and use Upload a plugin.

ChatGPT — download the same ZIP and ask Plugin Creator to create a private personal plugin from it; open the returned plugin page and install it. Your account needs Plugins / Plugin Creator support. Attaching a ZIP to an ordinary chat does not install a plugin. Full platform instructions.

All 34 skills through the CLI: Codex, Claude Code, Cursor

Requires Node.js 22.20+ and Git. Run the command for your client:

System	All marketing skills across your projects
Codex	`npx skills add LuCheremisina/marketing-skills --skill '*' -a codex -g -y`
Claude Code	`npx skills add LuCheremisina/marketing-skills --skill '*' -a claude-code -g -y`
Cursor	`npx skills add LuCheremisina/marketing-skills --skill '*' -a cursor -g -y`

--skill '*' selects every discovered skill; keep the quotes to prevent shell filename expansion. -a selects the client, -g installs across projects for your user, and -y skips prompts. For one project, omit -g and run from its folder. Avoid --all when targeting one client: it also selects all agents.

To choose skills and clients interactively:

npx skills add LuCheremisina/marketing-skills

One skill separately

npx skills add LuCheremisina/marketing-skills --skill che-deep-research

Select your client if prompted. This installs into the current project by default. For a global Codex installation, for example, add -a codex -g. You can also import the individual Deep Research ZIP through a supported Skills interface.

Ask your agent to install

Paste this into Codex, Claude Code or Cursor with terminal access:

Install all 34 marketing skills from LuCheremisina/marketing-skills
for your own client, globally. Use:
npx skills add LuCheremisina/marketing-skills --skill '*' -a <your ID: codex, claude-code or cursor> -g -y
Replace <...> with your client ID first. If Node.js or Git is missing,
explain what must be installed. If existing copies have local edits,
show the differences and preserve them first. After installation, check
the skill list and che-deep-research resources. Report what you verified.

Plugin and CLI installations are alternatives: avoid enabling duplicate copies. Local installation does not automatically carry into a cloud session; install in the environment running your agent. Frozen-release verification and rollback.

How to use

After installing, open a new chat or session. Enable the plugin in your installed list if needed. Select a skill in the client interface, or request it by name:

Use che-deep-research. Compare three marketing automation services for
a B2B company. Geography: Russia. Decision horizon: 12 months.
Include sources, a comparison table, limitations and a recommendation.
Ask for missing inputs first.

In Claude Code with the plugin installed, use its namespaced command:

/marketing-skills:che-deep-research Compare three marketing automation services for a B2B company

In Codex with CLI-installed Skills, you can explicitly select the skill:

$che-deep-research Compare three marketing automation services for a B2B company

In Cursor, use the skill picker or request it by name; in ChatGPT, select the installed plugin and describe your task. The plain-language “Use che-deep-research…” request also works in other supported systems; slash-command syntax is client-specific.

Supply your business context, goal, audience, geography, period, sources and desired output. You can attach a brief or point to a file. Advertising, Wordstat, Metrika and email workflows need separate connections or authorized exports; installing the skills does not create them. Connect your sources.

Check, update or remove

For CLI-installed Skills:

npx skills list -g
npx skills check
npx skills update che-deep-research -g
npx skills remove che-deep-research -g

To update all global CLI skills, including those from other sources: npx skills update -g. For a project installation, use npx skills list, npx skills update -p, and removal without -g from its folder. Preserve local edits before updating.

Manage native plugins through Plugins, rather than npx skills update. Plugin updates and removal. Check discovery of 34 skills, read one reference and try a sample task; file presence alone does not prove every workflow executes successfully.

Start here: 5 useful marketing skills

Your task	Skill and example	What you can get	Needs
Research a market or compare options	Deep Research	A sourced report with findings, uncertainty and decision options	A research question; web access or supplied sources
Audit search and AI-search visibility	SEO/GEO Auditor	A website audit with evidence and prioritised fixes	An authorised URL; Python and crawl/read tools
Find demand in Yandex search	Wordstat Keyword Research	Query clusters with region, frequency and intent	Wordstat connection or an authorised export
Diagnose advertising performance	Yandex Direct Analytics	Source-separated advertising and attribution tables	Direct/Metrika read access or prepared exports
Create a sourced expert article	Expert Article Workflow	A draft, claim ledger and editorial checks	A brief, sources and an author/brand profile

Each page includes a sample request, expected output, install command and individual ZIP. Full task catalogue →

Get help and share what worked

Installation questions · Request a skill · Share a result

Include your client, chosen skill and a synthetic example. Keep account credentials and client records private. If a workflow helped, star the repository to find it again; watch releases for published updates.

What's in the library

The packages follow the Agent Skills format. Marketing workflows are MIT licensed; engineering packages retain their original licences. Brand, business and author profiles are supplied at run time, so the public packages contain no client profiles or credentials.

Current published release: v2.0.7. Individual versions and sources are tracked in the catalogue. Verify downloads using the release manifest and SHA-256 checksums.

Choose a workflow

Business task	Start with	Reviewable result
Understand a market or customer decision	`che-product-market-intake`, `che-deep-research`, `che-research-loop`	Research scope, evidence, uncertainty and next-process decision
Assess campaign performance	`che-direct-analytics-skill`, `che-direct-analyst`, `che-audit-unisender-email`	Source-separated metrics, attribution limits and prioritized actions
Plan growth and monitor demand	`che-build-predictive-growth-dashboard`, `che-wordstat-trend-radar`, `che-seo-demand-analytics`	Scenarios, trend signals and their assumptions
Improve search and AI-search visibility	`che-seo-geo-auditor`, `che-seo-geo-auditor-pro`, `che-wordstat-keyword-research`, `che-topic-planner`	Contextual findings, intent and a practical content plan
Create sourced expert content	`che-expert-article-workflow`, `che-seo-article-generator`, `che-news-editorial-workflow`	Draft, evidence ledger and conditional integration handoff
Produce social and audio content	`che-content-carousel`, `che-telegram-content-creator`, `che-news-voiceover-script`, `che-short-video-production`	Storyboard or script; actual media only when the runtime is available
Prepare covers and apply brand guidance	`che-brand-profile-guide`, `che-editorial-cover-prompts`, `che-cover-production`	Brand-aligned concepts, prompts and validated image export
Monitor industry and AI news	`che-industry-news-digest`, `che-daily-ai-news-digest`	Sourced digest with time window, deduplication and delivery status
Accept a delivery or prevent a regression	`che-brief-acceptance`, `che-skill-regression-check`	Requirement evidence, unresolved gaps and regression verdict

See the complete catalogue for inputs, outputs and required connections. Cloudflare skills and autopilot are optional engineering support packages; original authorship and licenses are documented in third-party sources.

Use with your AI system

System	Distribution path / requirement
Codex	GitHub marketplace plugin or all / selected Skills through the CLI
Claude Code	GitHub marketplace plugin or CLI-installed Skills
Cursor	All / selected Agent Skills through the CLI
ChatGPT	Plugin ZIP where account import is supported; tools vary by account
Claude / Cowork	GitHub marketplace plugin, plugin ZIP or individual skill ZIP; account-dependent
Grok Bot	Saved skill library where supported; resources may need separate transfer

The platform guide describes requirements and limitations. Full execution across all platforms is not verified; test your selected workflow in your environment.

Use quick or verified installation for the route that fits your environment. Never place credentials or client data inside an installed skill. Detailed platform guidance and official sources are in PLATFORMS.md.

Evidence and quality

The library separates facts, hypotheses, recommendations and measured outcomes. Missing inputs or unavailable tools produce a clear limitation rather than fabricated research, revenue or delivery success. Research workflows preserve source, date, scope and uncertainty. Publication and external actions follow explicit user authorization.

Release preparation includes YAML parsing, resource and attribution checks, privacy scans, deterministic ZIP archives and SHA-256 manifests. Use the workflow quality checklist and skill-local synthetic fixtures to assess a selected workflow. These checks do not establish full native execution.

Search and AI-search recommendations use current primary documentation. FAQPage and llms.txt are not universal ranking requirements; indexing, citations and business impact must be measured separately. See Google AI-search guidance.

Versions and maintenance

Immutable release packages carry individual skill versions, source commits and checksums. Third-party upstream versions remain distinct from local adaptation versions. A changed package must receive a new version; filesystem dates are not used to decide which copy is newer.

Public upstream checks are scheduled monthly, on the 1st day at 00:00 Asia/Novosibirsk, with read-only repository permissions. Updates require review; there is no automatic merge, installation or release publication. The library may lag upstream between checks. See governance and changelog.

Methodology and attribution

Methodologist — Любовь Черемисина (Lyubov Cheremisina).

Original universal methodology workflows are licensed under MIT. Third-party packages retain their original authors, licenses and notices; see THIRD_PARTY_NOTICES.md. Sources without established redistribution permission are linked rather than republished. Methodologist attribution does not make her the author, narrator or brand of a client's output.

For citation use CITATION.cff. Contributions should include provenance, appropriate licensing, synthetic examples and evidence of the affected checks; see CONTRIBUTING.md.