agentleFS
kostja94kostja94/marketing-skills/skills/keyword-discovery/SKILL.md

keyword-discovery

Skill1.1k starsChanged 26 days agoMITView on GitHub

Finds long-tail search expressions from a seed, competitor sites, search suggestions and customer language. Use for keyword expansion and demand discovery.

2 other projects use a version of this.

Trending

More talk lately.thin dataabout the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 0 mentionsMar 2026: 1 mentionApr 2026: 0 mentionsMay 2026: 0 mentionsJun 2026: 0 mentionsJul 2026: 0 mentionsAug 2026: 0 mentionsSep 2026: 0 mentionsOct 2026: 2 mentionsNov 2026: 1 mention
Dec 2025Jun 2026Nov 2026
Mentions a month of the kostja94/marketing-skills repository that our weekly crawl found: only 4 in all, so this is a hint, not a trend. There is no adoption data for this file yet.

Most recently taken up by

The file

Files
SKILL.md

Keyword Discovery

Start with the seed, product boundaries, target market and language. Follow the user's highest-value use case first. Explore alternative wording, action, input, desired output, audience, use case, workflow scale and quality or failure conditions. Combine dimensions into plausible tasks and retain the original wording of observed queries. For another market, find local usage independently, including colloquial terms.

Build the base expression family before expanding features or scenarios. Separate the action, object and optional qualifiers. Probe natural verb, process noun, tool noun, word order and number variants with and without qualifiers. Check tool-family heads, cross-root task paraphrases and audience-specific language when relevant. Preserve the raw strings; removing a modifier can broaden intent. For generate song with ai, probe generate song, generate a song, song generation, AI song generation, song generator, AI song generator, song maker and make a song with AI. Query distinct root forms separately to discover branches that the original string misses. Keep unattested forms labeled as hypotheses.

For a narrow seed, identify the constraint that narrows it (such as a symptom, material, audience or use case). Offer the nearest useful broader query by relaxing that constraint while preserving the task. Show the parent followed by the input seed and explain the difference. Label a parent as observed wording or an inferred topic. Keep the original seed as the research scope. Use the parent to find relevant siblings when helpful. Multiple plausible parents can reflect different dimensions. A parent-child relationship alone does not justify targeting both on one page.

Inspect the target-market search results for competitors serving the task. Open relevant navigation, sitemap URLs, feature pages, industry pages and help topics. Check title, H1, actual inputs and outputs, and the referring page. A link from an ecommerce solution to a white-background tool suggests a product-photo hypothesis. Record the page URL and label the derived query as inferred. A competitor page supports coverage; claims about its traffic or rankings require separate evidence. Reuse competitor-research findings if available.

Probe Google Autocomplete with the seed and promising partial phrases. Capture People Also Ask questions verbatim and People Also Search For or Related searches under the label actually displayed. Record the triggering query, market, language, date and source for each observation. A featured snippet gives clues about answer format. Use relevant GSC queries, ad search terms, site search, support language or user-provided exports to extend coverage. Re-query valuable new branches separately. If ordinary search access times out or cannot expose these surfaces, use an available authorized cloud browser or personal agent to run the queries in the target locale. Supply seeds, locale and required surfaces; request verbatim entries, triggering query, displayed module label, date and source URL or screenshot. Distinguish a module absent on an inspected page from a page that could not be inspected. If no such session is available, return the pending collection request and continue with available evidence. Keep synthesized phrases separate from observed suggestions, even when they sound natural.

For background remover, distinguish removing a white background from adding white to a product photo. Explore image versus video, ecommerce versus passport, and preserving fine edges versus repairing an existing cutout. Use these as probes until observed.

Hand keyword-clustering a deduplicated candidate pool retaining raw wording, source or URL, observed/inferred status and branch. Surface unsupported product needs separately. Stop at saturation or the agreed boundary.

For creative generation, explore the attributes of the output as well as the inputs. Probe genre or style, occasion, audience and intended use. For a pattern such as [genre] song generator, sample contrasting values and check demand evidence individually. Keep unverified combinations as probes.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.