Use content-strategy to decide why the next piece deserves to exist, rather than simply generate titles. It works for a new blog, reprioritizing an existing library, or turning interviews and search exports into topics. Business goals and production capacity are essential inputs.
Install and invoke
Run this in your project terminal to install only this skill for Claude Code:
npx skills add coreyhaines31/marketingskills --skill content-strategy -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 content-strategy and supply your material. The installation command belongs in the terminal, not in your content brief.
Supply evidence before requesting ideas
Combine an article inventory, Search Console queries, customer questions, and your available weekly time. .agents/product-marketing.md supplies positioning, but actual questions still need evidence. You can begin without a keyword tool; do not ask the agent to invent search volume.
Example brief:
Use content-strategy to plan next month.
Read gsc.csv, support-questions.md, and articles.csv. We want design teams that need collaborative annotations.
Capacity is one article weekly and no video. Suggest three topic areas, then choose four pieces.
For each, give the reader question, evidence, intent, missing material, and distribution destination. Mark unknown data.
Check that each topic connects naturally to a problem the product addresses. Searchable content answers existing questions. Shareable content needs an insight, story, or useful material worth passing along. Discussion potential does not automatically establish buying intent. Map topics across learning, comparison, decision, and implementation stages instead of publishing only broad introductions.
Convert the ranking into work
The skill scores customer impact, content-market fit, search potential, and resource requirements. Ask for the basis of each score. Numbers are a decision aid, not a demand forecast. Check the direction of the resource score: a more expensive piece should not rise automatically because its numeric value is larger.
For the selected articles, specify questions to answer, relevant internal links, missing examples, and material that can become shorter social posts. Most blogs can keep ordinary article URLs and useful internal links. A topic cluster does not require rebuilding the entire URL hierarchy.
If the output remains a long list of similar titles, narrow the audience and time horizon, and remove duplicate intents. Give each piece a measurable job: resolve a usage obstacle, help compare alternatives, or support a resource download. Review search results, subsequent behavior, and reader feedback against that job. The number of published pages alone does not tell you whether the plan worked.
Turn a question bank into an executable editorial plan
Combine sales-call questions, support tickets, site search, forum discussions, and keyword data. Cluster them by customer problem before ranking by search volume. For every proposed piece, record buyer stage, the question it will answer, available evidence, appropriate format, and owner. A popular term with no distinctive answer can wait; a recurring question that your product can genuinely resolve may be a better first article.
Ask for three to five content pillars and related topics, then turn the highest-priority item into a writing brief: intended reader, what they already know, three questions to answer, usable source material, and next action. Searchable pieces should satisfy a clear information need; shareable pieces may come from original observation or a useful point of view. Measure them accordingly. The official 60/30/10 searchable, shareable, and experimental split is a starting framework, not a quota that a tiny team must satisfy each week.
Place only feasible pieces on the calendar. Confirm an expert, case, or dataset is available before scheduling a claim-heavy article. If it is not, make “interview two customers” the next task rather than asking an agent to invent their experience. Review published work monthly for relevant visits, changed product questions, and pages that need updating before choosing the next batch.
Example: turn support questions into an implementation guide
Suppose more than a dozen tickets this month ask why project members cannot see client comments. Supply redacted tickets, the current help pages, and permission documentation. Ask: “Use content-strategy to decide whether this deserves an implementation-stage guide. Identify the three most common misunderstandings, the documented permission rule behind each, an article outline, and questions for the product team. Show which existing help pages should link to it.”
A useful plan puts the guide in the right topic cluster and makes clear that it solves a permission task, not merely targets a keyword. If the agent proposes a separate blog post for every ticket, merge repeated questions. If it relies on an unreleased feature, return to product facts. Once the guide is published, inspect whether similar tickets still arrive and whether users can find it from related help pages before expanding the topic.
Two inspectable editorial outputs
In a hands-on video, marketer Suman Chatterjee supplied Tiiny Host's audience, content goals, capacity of two posts a week, and competitors, then connected DataForSEO. The report included keyword clusters and search volume; a separate Tiiny Host Skill published it as a shareable page. Keep the data source and date in your own topic sheet. Search volume comes from the connected service, not from content-strategy itself.
Phoebe Fu's public artifact has four editorial pillars, eight prioritized ideas, and one hub article with five narrower supporting questions. Each idea includes a stage, format, and planning rationale. The scores are editorial judgments, not measured keyword demand. Without search data, this sort of backlog helps decide what to investigate next; it is not a traffic forecast.