Use marketing-ideas when you are unsure what to try next. Its categorized library is useful for exploring options, but expanding every idea is rarely a good first task. Aim for three to five candidates that fit your audience, stage, and capacity.
Install and invoke
Run this in your project terminal to install only this skill for Claude Code:
npx skills add coreyhaines31/marketingskills --skill marketing-ideas -a claude-code
Node.js and npx must be available. For Codex, replace -a claude-code with -a codex. In the agent conversation, explicitly ask it to use marketing-ideas and supply your material. The installation command belongs in the terminal, not in your content brief.
Explain the constraints first
Use .agents/product-marketing.md for positioning, then add the immediate objective, weekly time, and available writing or development skills. Include unsuccessful past attempts so the agent does not simply repeat them.
Example brief:
Use marketing-ideas to suggest three acquisition experiments for this month.
We sell a design-review tool. One person has six hours weekly, no advertising budget, and no established community.
We published eight blog posts; results are in review.md. Give the fit, first three steps, required material, and observation metrics for each idea.
Choose one experiment to begin this week and explain why the others should wait.
Inspect how each recommendation reaches relevant people. Content, partnerships, product sharing, and free tools solve different problems; their presence in a library does not justify doing all of them. Free channels consume time, and a quick action does not guarantee quick customers. Make starting conditions explicit.
Turn an idea into a testable action
Define a concrete artifact and audience for the selected direction. “Create content” should become an article answering a particular question. “Find partners” should name the type of complementary business and the shared value for readers. The first step must be executable now, not an aspiration such as building influence.
Set a review period and specify signals that would justify continuing, adjusting, or stopping. Without a baseline, begin by checking whether the process works and whether relevant users respond. Do not fabricate conversion targets or treat scores as forecasts. Put external data, paid services, and partner cooperation in the dependency list.
Use the broad library to discover omissions, not to populate every category. Once you choose an experiment, hand execution to the appropriate content, email, tool-planning, or launch skill. Bring back artifacts, feedback, and actual effort for the next review. Generating more ideas without completing any test rarely reduces uncertainty.
Use resource constraints to discard attractive but impractical ideas
This skill is an idea library; it cannot infer the team’s capacity. Supply company stage, one growth objective, weekly hours, budget ceiling, and channels already tried. Ask it to select three candidates instead of listing thirty, and for each explain why it fits now, the first two actions, required people, and a signal to stop. A solo operator should explicitly exclude ideas that depend on a designer, sales team, or a large stream of fresh content.
Frame an idea as a one- or two-week test. For example, put a downloadable checklist answering a repeated customer question on an existing help page, then record whether relevant readers open and use it. The expected outcome should be an observable behavior, not an invented forecast such as “500 leads.” When comparing channels, keep the audience, time window, and measurement definition clear; unlike exposures are not equivalent.
At the end, keep a one-page record of effort spent, meaningful behavior seen, new user questions, and whether the test merits expansion. A failed idea can narrow your options. One quiet post does not prove an entire channel is ineffective. Feed this record into the next recommendation instead of repeatedly asking for another hundred generic marketing ideas.
Example: choose a two-week experiment before a trade event
A small team will attend an industry event next month. It cannot fund a full brand campaign but can interview existing customers. Give the skill the event audience, permission to use customer material, available hours, and current assets. Ask: “Use marketing-ideas to suggest three actions we can finish before the event. State preparation, dependencies, and a signal visible within two weeks; choose only one to pursue. Do not turn a private interview into a public endorsement.”
Check whether the chosen action fits the hours available. A one-page answer to common attendee questions may be more executable than simultaneously building a new website, podcast, and video series. If the only success measure is “exposure,” ask whether anyone raises a relevant question, books a follow-up, or actually uses the resource. Record effort and feedback after the event before repeating the tactic.
What public examples demonstrate
An independent multi-brand playbook allocated marketing-ideas to weekly micro-actions rather than treating the idea list as a complete marketing plan. That is a planning approach, not a completed campaign result.
Developer siegerts describes a bounded firsthand use: in Kiro CLI he installed Corey's pack for the fictional SurveyApp demo site, used marketing-ideas for content and positioning suggestions, then combined seo-audit and copywriting to review pages. The resulting work list covered social metadata, titles and descriptions, CTAs and FAQ. He stresses checking suggestions against product facts. This illustrates how to organize edits, not real customers, conversions or growth.