Living collection
Research and sources | Marketing skills for Claude Code and Codex
Marketing skills help AI agents work on audience research, landing page copy, SEO and commercial messaging. This collection compares skills and related resources for Claude Code, Codex and other agents by the task they support, the data they need and their documented compatibility.
For research, Last 30 Days gathers recent conversations, Watch analyses videos and Firecrawl extracts information from web pages. For copywriting, SEO and conversion work, Marketing Skills by Corey Haines brings specialised instructions together in one package. Each profile distinguishes agent skills from services and tools that can be connected to an agent.
Start with the job: understand your customers, prepare a landing page or review a campaign. Then compare setup, cost, sources and limitations. Resources link to a detailed review when one is available, so you can judge whether they fit your workflow.
10 resourcesOrganized by use
Research and sources
Research and sources
6 resourcesFull profiles and sources
Last 30 Days
Looks for current conversations where they happen and separates signals, opinions and noise before research.
Best for: Investigating recent discussions and signals with links to their sources.
Why it belongs here. Helps an agent find recent public conversation and keep a route back to each source.
Its limit. Coverage depends on available platforms and access. A recent sample is not a representative survey.
It helps you hear what people are saying now and find the original posts. Check what it could not reach before drawing a conclusion.
Compatibility and setupRequirements and agent differences
Setup. Check the skill's current agent setup and platform access requirements in its repository.
Real cost and licenseResource, agent and extra costs
Practical cost. The skill may rely on services with their own limits or access terms; check the sources needed for your research before running it.
Example and alternativeWhen it helps and what else to use
Useful route. Use it to collect recent questions about a market and follow the linked posts before summarizing recurring themes.
Sources and reviewWhat is documented and what was tested
Evidence. This entry uses the existing Marketing Ultra review and the project's official source. Author documentation is not described as a hands-on test by Dani.
Resource detailsEditorial profile and links
Watch
Breaks a video into frames, audio and a transcript so an agent can examine what it actually contains.
Best for: Investigating video content through frames and transcripts.
Why it belongs here. Brings visual and spoken evidence together when a video is the source material.
Its limit. Frame sampling can miss brief details and transcripts can contain errors; check decisive moments against the original video.
It gives the agent a set of frames and words from a video. You still need to check whether those samples capture the part that matters.
Compatibility and setupRequirements and agent differences
Setup. Check the repository's setup and required video tools for the agent and operating system you use.
Real cost and licenseResource, agent and extra costs
Practical cost. The workflow may require local video and transcription tools or external services. Check those requirements before treating the skill as a free one-step operation.
Example and alternativeWhen it helps and what else to use
Useful route. Use it to inspect a video claim or extract evidence for research, then return to the original timestamp before citing a detail.
Sources and reviewWhat is documented and what was tested
Evidence. This entry uses the existing Marketing Ultra review and the project's official source. Author documentation is not described as a hands-on test by Dani.
Resource detailsEditorial profile and links
Firecrawl
A web extraction API that turns pages into cleaner Markdown or structured data for agents and language models.
Best for: Giving a coding agent access to current web content.
Why it belongs here. Provides reusable web extraction when an agent needs to work with pages and external data.
Its limit. Extraction can miss or distort content. Check the result and the site's terms before relying on it.
It helps an agent read web pages in a usable format. It does not decide whether the extracted information is correct.
Compatibility and setupRequirements and agent differences
Setup. Firecrawl is an API service, not an interchangeable skill. Verify the integration and credentials required by your agent.
Real cost and licenseResource, agent and extra costs
Practical cost. Hosted usage and self-hosting have different plans and maintenance costs. Check current pricing and infrastructure requirements.
Example and alternativeWhen it helps and what else to use
Useful route. Use it for a repeatable collection of pages; for a single lookup, first consider the agent's existing browsing tools.
Sources and reviewWhat is documented and what was tested
Evidence. This entry uses the existing Marketing Ultra review and the project's official source. Author documentation is not described as a hands-on test by Dani.
Resource detailsEditorial profile and links
Claude SEO
Full SEO kit for Claude Code with 18 agents that prioritize findings and add falsifiability to every recommendation. MIT, no cost per domain.
Why it belongs here. Full SEO kit for Claude Code with 18 agents that prioritize findings and add falsifiability to every recommendation. MIT, no cost per domain.
Its limit. The SKILL.md injects a promotional footer for the author's community into generated reports. The 0–100 score is an internal heuristic, not a Google signal.
Give it a URL and Claude Code returns an SEO report with issues sorted by priority and a concrete way to check whether each fix worked.
Compatibility and setupRequirements and agent differences
Compatibility. Claude Code on macOS, Linux, and Windows. Requires Python 3.10+. Codex SEO is a separate port outside this package.
Real cost and licenseResource, agent and extra costs
Cost. The skill is free (MIT). You need an active Claude Code subscription. Since v2.3.1, five agents run on Opus, which increases consumption. Optional extensions (DataForSEO, Ahrefs) require separate paid accounts.
Example and alternativeWhen it helps and what else to use
Example. /seo audit https://your-site.com for a full audit with up to 18 agents. /seo page for a single URL. /seo schema for markup.
Alternative. Screaming Frog for deep crawling, Ahrefs Site Audit for link analysis with its own index, Semrush Site Audit for competitive visibility.
Sources and reviewWhat is documented and what was tested
Review. Data from the GitHub repository, official website, and public sources (Reddit, YouTube). The tool was not run for this review. September 2026.
Resource detailsEditorial profile and links
OpenClaudia Skills
It bundles more than 70 marketing skills into an open package and provides a specific route for querying GA4 reports and conclusions from a coding agent.
Why it belongs here. It bundles more than 70 marketing skills into an open package and provides a specific route for querying GA4 reports and conclusions from a coding agent.
Its limit. The skill does not include automatic Analytics access: it requires Google OAuth, property access, and human verification of reports; nor is there a documented independent execution.
In plain English: it lets you ask an assistant to query GA4 from the terminal and organize the data, provided you first connect your Google account correctly.
Compatibility and setupRequirements and agent differences
Compatibility. Claude Code is the main documented requirement. Codex supports a specific path for skills; Grok CLI, Gemini CLI, Cursor, and Windsurf can use the content as instructions or rules, without a guarantee of identical operation.
Real cost and licenseResource, agent and extra costs
Cost. The package is free and MIT-licensed. Credentials, permissions, and any external services needed for other skills fall outside that free offering.
Example and alternativeWhen it helps and what else to use
Example. Install google-analytics, configure OAuth, and request a basic report from an authorized property, then compare its figures with GA4.
Alternative. Keeping analysis in the original SaaS dashboards avoids adding another layer of instructions and credentials, although it requires switching interfaces for each task.
Sources and reviewWhat is documented and what was tested
Review. Profile reviewed on September 25, 2026, based on the OpenClaudia Skills repository and main documentation.
Resource detailsEditorial profile and links
GEO SEO Claude
It brings citability checks, crawler access, schema, content and GEO reports together in a single Claude Code skill.
Why it belongs here. It brings citability checks, crawler access, schema, content and GEO reports together in a single Claude Code skill.
Its limit. Its scores describe readiness through heuristics; they do not prove real mentions or citations in AI assistants.
In plain English: it checks whether your website is organised so AI search engines can crawl and understand it, but it does not confirm that they are showing it.
Compatibility and setupRequirements and agent differences
Compatibility. The documentation requires Claude Code CLI, Git and Python 3.8 or later. Windows requires Git Bash; there is no verified compatibility with Codex, Cursor or other agents.
Real cost and licenseResource, agent and extra costs
Cost. The repository publishes the code under an MIT licence. Claude Code, external services and the commercial community linked by the author have independent terms; no single price in euros has been verified.
Example and alternativeWhen it helps and what else to use
Example. It can review robots.txt, structured data and content blocks before preparing a GEO report for a client.
Alternative. The dossier does not establish a direct replacement. Claude SEO is a related tool that overlaps with general technical SEO auditing.
Sources and reviewWhat is documented and what was tested
Review. This listing was reviewed on 27 September 2026 using the repository, its README and verified licence information. The installation was not run.





