How to use the Competitive Teardown agent skill

What is an agent skill, and what does this one make the agent do?

An agent skill is a folder containing a SKILL.md file and supporting files that an AI agent can load when a task matches the skill's description. Competitive Teardown is one such skill. Once it is installed, the agent uses it when you ask for a competitive analysis.

The skill's stated goal is to answer the question "how do we differ from these competitors, and what should we do next?" The report it produces includes a header with market, product category, report date and a one-line conclusion; an executive takeaway of three judgments, each with a "so what"; a positioning map; competitor cards; a feature comparison matrix; a pricing and packaging read; UX and messaging notes; three opportunity windows; and 30, 90 and 180-day recommended moves.

The skill picks one of three HTML visual styles based on the input: a light war-room grid (the default), a dark radar map for security, AI, developer tools or platform competitors, or a paper analyst dossier for research notes and investment-style memos. It does not mix styles. Where your material is thin, the report lists evidence gaps rather than inventing numbers.

Install Competitive Teardown in Claude

Download the skill ZIP from this page and unzip it. You can install it in one of two ways.

For a personal install, unzip it into ~/.claude/skills/ so the skill is available across your projects. For a project-only install, unzip it into the .claude/skills/ folder inside that project. The skill folder should contain SKILL.md along with the assets/ folder, example.html and example.md.

Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.

• Personal: ~/.claude/skills/
• Project: .claude/skills/ inside your project folder
• Apps: Settings → Capabilities → Skills (upload)

Check the requirements first

The source lists no special requirements. It does not run scripts and does not use the network, so the skill works only with the competitor material you supply. It does not browse competitor websites or check claims against live sources.

The main requirement is data. The skill is built for 2 to 6 competitors. If you give it only one, it produces a single-competitor deep dive. If you give it no competitor data, the report will be mostly evidence gaps.

• No listed prerequisites; the source lists none.
• Does not run scripts.
• Does not use the network.
• Input must come from you: pricing pages, changelogs, reviews, sales notes, interview notes, screenshot descriptions or links you paste.

How to trigger the skill with a natural request

You do not need a special command. Describe the task in plain language and include the competitor material. The skill's description covers positioning maps, feature matrices, price comparisons and opportunity windows, so requests that mention these or a competitive review should match it.

Name the competitors and paste or attach the notes. If you want a specific visual style, say so; otherwise the agent chooses one based on the input. Make sure your request asks for a report rather than an article or a product requirements document, since the skill is built for decisions, not prose.

• "Make a competitive teardown of Acme, Beta and Gamma using the notes below."
• "Compare our product against these three rivals and give me a positioning map and feature matrix."
• "Turn these pricing pages and user reviews into a strategy report with 30/90/180-day moves."
• "Use the radar style for this: here are the security tools we are up against."

Example: what a good input and output look like

The bundled example, example.md, is a teardown of three AI meeting assistants written for a fictional product called LoopNote. Each competitor has its positioning, target user, pricing signal, strength, weakness and messaging recorded as short bullets. This is the level of detail that lets the report place competitors on a map and fill a feature matrix.

The example's competitor entries show the kind of signals the skill works from. For instance, one entry notes a free plan plus paid tiers with a team orientation, and another notes a public site that emphasises simplicity. The skill turns observations like these into pricing reads and opportunity windows.

• Good input: a positioning line, target user, pricing signal, strengths and weaknesses for each rival, plus quoted messaging.
• Weaker input: a single competitor name with no source notes. Expect many evidence gaps.

Safety notes and limits

The skill only uses the material you supply. It does not verify facts, pricing or claims against live sources, so treat its output as a working draft for discussion. Do not rely on it as verified fact for legal decisions or investment decisions.

The skill is also instructed not to invent market share, ARR, customer names or pricing figures. Anything missing should appear as "not found in source" or "unknown". Check those markers before you present the report.

• Verify prices and claims against the live competitor sites before a decision.
• Do not use the output as verified fact for legal or investment decisions.
• Expect gaps if your input is thin; add notes instead of trusting invented detail.

Tips for better reports

Give the skill sources it can compare on the same dimensions: price, target user, core promise, strengths and weaknesses. Consistent notes make the feature matrix and positioning map more useful. Quoted messaging and specific changelog items are especially useful for the UX and messaging notes.

Read the evidence gaps section first. It tells you what to research next, which is often more valuable than the recommendations themselves.

• Keep to 2 to 6 competitors for a focused matrix.
• Include your own product in the notes so the "Us / Opportunity" column has content.
• Choose the style deliberately: war-room for general product teams, radar for technical markets, dossier for formal research memos.