Build a Competitive Analysis Claude Skill for PMs
Competitive teardowns are the work PMs always say they'll do next quarter. A competitive analysis Claude Skill is how it actually gets done this quarter, in an afternoon, and how it stays current after that.
Why Competitive Analysis Is the Most-Skipped PM Workflow
Competitive analysis is high-value, low-urgency. That's the worst possible combination on a PM's calendar.
Nothing breaks tomorrow if you don't do it. No engineer is blocked. No customer call depends on it. The roadmap doesn't shift this week because you haven't read Competitor X's latest changelog. The work always gets pushed by something more immediate.
But the cost of skipping shows up later, and the bill compounds. You walk into a sales call without a clean answer about Competitor X's pricing. You ship a feature that overlaps with what they launched six months ago, and your team didn't notice. A board member asks about competitive positioning and you scramble through a stale doc nobody updated.
Right now, the urgency curve has shifted. Products are moving faster. AI-first competitors are launching every month. The teardown you wrote 18 months ago describes a market that doesn't exist anymore. The work was always worth doing. Now it's also overdue.
What a Great Competitive Teardown Includes
A useful competitive teardown is five sections, not a wall of bullet points.
Feature comparison. What does each competitor offer that you don't, and what do you offer that they don't? The honest version, not the marketing version. Their pricing page tells you what they sell. Their changelog tells you what they're investing in. Look at both.
Pricing matrix. Current pricing tiers, packaging, and any quotes from their actual pricing page. Pricing changes more often than anything else in a competitive teardown, which is why a static analysis stales fast and why this section needs the most discipline.
Positioning analysis. How does each competitor describe what they do, in their own words? Pull quotes from their hero copy, their about page, their G2 listing. The pattern of words they choose tells you what category they're targeting and what objections they're handling.
Strengths and weaknesses. Strengths are easier (they say them out loud). Weaknesses come from G2 reviews, Reddit threads, and the support tickets your sales team forwards you. The honest weaknesses are the ones that show up across multiple sources, not the ones that show up only in your team's hopes.
Strategic implications. Which competitor is closest to you? Which is moving fastest? Which is a category-level threat? The first four sections are the data. This section is the competitive intelligence layer: what you do with the data, and which strategic moves it implies for your own roadmap.
Building a Competitive Analysis SKILL.md
Three files in one folder. SKILL.md holds the instructions and trigger. MATRIX.md holds your team's teardown format (the five sections in your team's preferred layout). PREVIOUS.md is the last teardown the Skill produced, so the next run can surface what changed.
Here's a working SKILL.md to copy and adapt:
The trigger description for AI competitor research should cover the surface phrases your team uses. "Teardown [competitor]," "compare X vs Y," "update our comp matrix," "what's [competitor] doing now." The Skill needs to fire reliably in every form the question shows up, not just when you remember the exact prompt.
Pairing the Skill with Live Web Browsing for Accuracy
Here's the part most AI competitor research tutorials skip. A Skill running on static training data is useless for competitive analysis.
Pricing changed last week. The feature comparison in your head is six months out of date. Competitor Y just announced a new tier yesterday. None of that is in any AI model's training data, no matter how recent the model is. A claude competitive teardown produced without live data is a competitive teardown of last year's market.
Live web browsing is what fixes this. In Claude Code for product managers, Claude can fetch competitor pages directly, read pricing, parse changelogs, and quote positioning verbatim. The Skill prompts Claude to use that capability for every claim that depends on current data.
The discipline is in the rules section of your SKILL.md. Never cite pricing without a source URL and the date it was checked. Always quote positioning verbatim, not paraphrased. If a page returns 404 or paywall, surface that as a gap rather than guessing. Without those rules, the Skill will happily produce a polished teardown that's quietly inaccurate, which is worse than no teardown at all.
Keeping Your Teardown Current Without Doing It Twice
The first teardown is the slow one. Every one after it should be a delta.
The pattern looks like this. Run the Skill once. Save the output as PREVIOUS.md in your project folder. Three months later, run the Skill again. Tell it to compare the new output to PREVIOUS.md and surface what changed. New pricing tier. A feature that shipped. A repositioning of their hero copy. A weakness that's now resolved.
The delta output is shorter and faster to produce than a fresh teardown, and it's actually more useful. You stop re-reading what you already knew and start tracking what actually moved. The teardown becomes a living document, updated quarterly, and competitor monitoring stops being a separate project you keep forgetting to schedule. The same Skill that produced the initial analysis keeps it sharp.
This is also where voice-of-customer evidence for competitive positioning starts paying off. The strongest competitive insights come from real customer language about why they switched, what they almost picked, and what they wish each competitor did better. A Skill that knows where to pull customer language from produces teardowns that hold up in front of leadership, not just in front of your slack channel.
Where to go from here
A competitive analysis only matters if it shapes your roadmap. The output of a teardown should feed back into your prioritization and your strategic bets, not sit in a Notion doc nobody opens. The next compounding step is running the teardown output through your other Skills so the competitive picture is built into how you sequence what to build next.
Frequently asked questions about building a competitive analysis Claude Skill
What does a competitive analysis Claude Skill actually produce?
A structured teardown with five sections: feature comparison, pricing matrix, positioning analysis, strengths and weaknesses, and strategic implications. Each section is grounded in current data: pricing from competitor pricing pages with source URLs, positioning quoted verbatim from their own copy, weaknesses cross-referenced from review sites. The output is a Markdown document you can drop into Notion, share with sales, or use as the basis for a roadmap discussion.
Can a claude competitive teardown handle multiple competitors at once?
Yes, and that's most of the value. Paste in a list of three to five competitor names or URLs, and the Skill produces a comparative teardown across all of them in one pass. For larger competitive sets (10+ competitors), batch them into related groups. Direct competitors in one run, adjacent-category competitors in another. The Skill produces clearer output when the competitors are roughly comparable.
Why does AI competitor research need live browsing?
Because pricing, packaging, and feature sets change every few weeks. A model trained six months ago does not know that Competitor X added a new tier last Tuesday. Without live browsing, the Skill produces a teardown that looks polished but is quietly out of date. Live browsing in Claude Code lets the Skill pull current pricing pages, parse changelogs, and quote positioning verbatim. Skip this step and the teardown is worse than not having one.
How do I keep a competitive teardown current without redoing it every quarter?
The delta pattern. Run the Skill once and save the output as PREVIOUS.md. Next quarter, run the Skill again and tell it to compare to PREVIOUS.md. The output is a list of what changed: new pricing, new features, repositioned messaging, resolved weaknesses. The delta run takes a fraction of the time of a fresh teardown and produces sharper insights because it shows movement, not just snapshots.
Should the Skill quote competitor pricing verbatim, or paraphrase?
Verbatim, every time. Paraphrasing is where bias and errors creep in. "Their starter plan is around $50/month" is the kind of phrase that ends up wrong in front of an exec. "Starter plan: $49/month, billed annually, includes 3 seats" with a source URL and a date checked is the version that holds up. Build the verbatim rule into your SKILL.md and refuse to ship teardowns without it.
