Marketing Skills
Measure Marketing Skills by accepted work and reader outcomes, not page count
Design a controlled pilot that separates model cost, editorial effort, citation observations and successful learning.
What you will learn
- Measure accepted output including correction time.
- Citation and recommendation need separate counters.
- Reader-task completion is more useful than article count alone.
Before you start
- Basic Markdown and repository navigation
- An owned page and a factual product brief
Explain the chapter’s workflow boundary and check its evidence or proposed acceptance task.
Key takeaways
- Measure accepted output including correction time.
- Citation and recommendation need separate counters.
- Reader-task completion is more useful than article count alone.
Count the resources the skill itself does not bill
The repository’s MIT license is not a zero-cost workflow. A host agent may consume model tokens; evidence collection may use paid services; integrations may need subscriptions. Review and correction also take time. Record these separately so a faster first draft is not mistaken for a cheaper accepted article.
For a controlled comparison, give the same factual brief and task to a baseline workflow and a skill-assisted workflow. Keep model, evidence access and acceptance criteria comparable. Record retries and corrections, not just the first response latency. We did not run that experiment, so this chapter provides a measurement design rather than a claimed productivity gain.
Keep visibility metrics distinct
A retrieved page, a cited URL, a brand mention and a product recommendation are different observations. The inspected ai-seo text explicitly separates citation from recommendation and asks for repeated observations because generated answers vary. An article can be cited for a definition without the publisher being recommended as a vendor.
Use a fixed query set, language, platform and collection date. Preserve sample size and the answer evidence for each observation. Three citations in five invented trials equal 60 percent arithmetically; that small illustrative ratio does not estimate your site’s long-term visibility or prove an improvement caused by a wording change.
Make learning an acceptance criterion
For a repository guide, ask a simulated beginner to locate prerequisites, identify the first expected result and explain one limitation. Ask an operator to locate rollback and security boundaries. These checks expose missing information that a keyword checklist can miss. Simulated readers are an editorial tool, not a substitute for recruiting actual readers.
Report editorial scores as self-assessment with reasons and unknowns. Leave search volume, campaign conversions and acquisition cost empty when they were not measured. This is particularly important when scaling to many projects: a high generated score cannot compensate for an article that never inspected the implementation it claims to explain.
Implementation steps
- 1
Define the same task for baseline and skill-assisted runs.
- 2
Record model, tools, revisions, retries and review time.
- 3
Retain observations and denominators for visibility metrics.
- 4
Label simulated reading and unknown business results.
Copy-ready example
// Invented observations, not site analytics.
const trials = [true, false, true, true, false];
const cited = trials.filter(Boolean).length;
const rate = cited / trials.length;
console.log({ cited, n: trials.length, rate }); // 3, 5, 0.6Frequently asked questions
Does 3/5 prove an SEO improvement?
No. It is a small observation ratio; attribution and generalization require an appropriate comparison and more evidence.
Is an editorial score a user-study result?
No. Label simulated personas and self-review clearly, and retain the reasons rather than presenting the score as independent validation.
Sources
- README.mdSource checked 2026-09-08
- LICENSESource checked 2026-09-08
- .claude-plugin/plugin.jsonSource checked 2026-09-08
- .github/scripts/sync-skills.jsSource checked 2026-09-08
- .github/workflows/sync-skills.ymlSource checked 2026-09-08
- .github/workflows/validate-skill.ymlSource checked 2026-09-08
- scripts/sync-partners.mjsSource checked 2026-09-08
- skills/product-marketing/SKILL.mdSource checked 2026-09-08
- skills/ai-seo/SKILL.mdSource checked 2026-09-08
- skills/seo-audit/SKILL.mdSource checked 2026-09-08
- skills/programmatic-seo/SKILL.mdSource checked 2026-09-08
- skills/ai-seo/evals/evals.jsonSource checked 2026-09-08
- tools/PARTNERS.mdSource checked 2026-09-08