Use competitor-profiling to understand several products’ positioning, pricing, and content direction. Its output is a dated research dossier rather than a ready-made comparison article. The full workflow uses separate scraping and optional SEO services; installing the skill does not grant those capabilities.
Install and invoke
Run this in your project terminal to install only this skill for Claude Code:
npx skills add coreyhaines31/marketingskills --skill competitor-profiling -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 competitor-profiling and supply your material. The installation command belongs in the terminal, not in your content brief.
Begin with a bounded scan
Supply two or three competitor URLs and the dimensions that matter. A first pass can focus on homepages and pricing: audience, value proposition, plans, and trial conditions. Use .agents/product-marketing.md to explain your own product and keep the investigation relevant.
Example brief:
Use competitor-profiling for a quick scan of competitors.txt.
Read homepages and pricing only, focusing on small-team fit and billing.
Identify available scraping and SEO tools first. Leave inaccessible metrics blank.
Save dated raw material, create one dossier per company and a cross-company summary, and separate facts from inferences.
The official workflow uses Firecrawl for discovery and extraction and DataForSEO for search metrics. These services need their own client integration. Without them, the agent can analyze supplied pages, but should not describe that as the complete automated workflow. Bound the scope and budget of paid data requests.
Check that conclusions lead back to evidence
Raw material belongs under competitor-profiles/raw/competitor/date/, separated into page, review, and SEO records. Synthesized dossiers sit in competitor-profiles/, with _summary.md for comparisons. Preserve previous date folders during updates so changes can be investigated.
For prices, record currency, billing period, seats, and usage limits. Label a customer-count statement as a company claim. Search estimates cannot prove actual customer counts. DataForSEO’s ETV is calculated from keyword search volume and click-through assumptions; figures from different locations or languages are not directly comparable.
Spot-check three conclusions against their original page and collection date. Unsupported team size, market share, or weaknesses should become unknowns or explicit inferences. Explain the differences behind proposed opportunities and list anything requiring hands-on product use separately. Reading a website is research evidence, not a product test.
Build a comparable profile that can be updated
Use the same fields for every competitor: positioning, intended customer, principal capabilities, pricing, proof, recent product changes, and implications for your product. Attach a URL and access date to important facts. “Apparently cheaper” belongs in an unknowns column until checked. Pricing changes quickly; an old screenshot is not a current term. If nobody has used the product, label the result a public-site scan and do not rate its lived user experience.
Start with one competitor. Read the homepage, feature and pricing pages, and changelog, then request a quick scan: “Profile example.com. Use only the pages I supply and public pages you can actually open. Separate observed features and prices from your interpretation and unknowns. List what a deeper profile would require.” Firecrawl or DataForSEO can supplement research when connected; absent those tools, the agent must not imply it obtained scraped pages or ranking data.
For each comparison, ask who would care about the difference and when the competitor might actually be the better choice. On the next update, check pricing and product changes first and record what changed. That produces a profile useful for positioning and sales questions, rather than a one-off collection of screenshots.
Example: treat hidden pricing as unknown
Suppose you need a profile of an enterprise SaaS competitor whose site says only “contact sales.” Supply its homepage, feature pages, changelog, and your own product’s positioning. Ask: “Use competitor-profiling for a quick profile. Record verifiable audience, capabilities, and plan names. Mark all prices unknown and list the questions a sales inquiry would need to answer. Finish with differences our product can substantiate and questions we cannot resolve yet.”
Open each cited source when reviewing. If the agent turns a public trial allowance into an enterprise annual price, or concludes the unpriced plan is “more expensive,” return that assertion to the unknowns column. If it proposes Firecrawl or DataForSEO, first confirm those services are actually connected. Otherwise use only pages that were opened. A later deep profile may add real product trials; a website scan must not be described as first-hand use.
What a third-party adaptation adds
UnifAPI's independent adaptation separates a quick scan from a deeper dossier, keeps dated source links for important fields, and cross-checks website, search, and public social signals. It demonstrates a maintainable research method, not a completed competitor report or business outcome. On the next update, recheck changing fields such as price, features, and recent releases before reusing an old conclusion.