product-teardown · diff
git:20260807.975fe85 to git:20260807.fac9f85
49 added, 90 removed. Audit A to A.
---
name: product-teardown
- description: Analyze a product or competitor across positioning, proof, copy, UX, design, pricing, and onboarding. Use when extracting evidence-backed lessons for another product.
+ description: Produce evidence-backed product or competitor teardowns and transferable principles. Use when analyzing positioning, copy, UX, onboarding, pricing, trust, design, or visible retention loops.
license: MIT
---
# Product teardown
- Analyze a product in context and turn observations into applicable lessons. Separate visible evidence from inference, and never copy protected expression, brand assets, or distinctive implementation.
-
- ## How to start
-
- Ask TWO questions before doing anything else. Ask them together.
-
- > "What's the product you want to tear down? And what are you building — so I know what to focus the 'steal this' section on?"
-
- Wait for both answers. If they give you a product but no context on what they're building, ask before analyzing. A teardown without a target is just criticism.
-
- If they haven't told you what specifically they want to learn — positioning, copy, UX, social proof, onboarding — ask that too. It sharpens the analysis.
-
- ---
-
- ## Phase 1: Get the Target
-
- The user provides one of:
- - A URL — fetch it and analyze the page
- - A product name — search for it, pull the landing page
- - A screenshot — analyze what's visible
- - A description — work from what they describe
-
- If they provide a URL, read the full page. If the product has multiple pages (pricing, features, about), read those too.
-
- Ask: **"What are you building?"** — The teardown is only useful if you know what to apply it to. If the user hasn't mentioned their own project, ask.
-
- ## Phase 2: The 7 Lenses
-
- Analyze through each lens. Not everything applies to every product — skip what's irrelevant.
-
- **1. First Impression (5-second test)**
- What do you understand about this product in 5 seconds? Can a stranger tell what it does, who it's for, and why it matters? If not, what's missing?
-
- **2. Positioning**
- How does this product position itself? What category does it claim? What's the implicit "us vs. them"? Who are they NOT for? The best products are polarizing — they pick a side.
-
- **3. Copy Quality**
- Score the headline, subheadline, CTAs, feature copy, and microcopy. Use the same rules as the landing-copy skill:
- - Headline under 10 words?
- - Specific benefit over vague promise?
- - CTAs pass the "I want to ___" test?
- - No "powerful", "seamless", "revolutionary"?
- - Does it sound like a human or a marketing team?
-
- **4. Social Proof**
- What proof do they show? Numbers, testimonials, logos, case studies, press mentions? Is it specific ("847 teams use this") or vague ("trusted by thousands")? Is it above the fold?
+ Analyze a product in context and convert observations into transferable principles. Do not copy protected expression, brand assets, or distinctive implementation.
- **5. Objection Handling**
- What reasons would someone have NOT to buy/use this? Does the page address them? Look for: pricing concerns, trust issues, competitor comparison, "is this for me?" doubts. Unaddressed objections are conversion leaks.
+ ## Choose a mode
- **6. User Experience**
- If it's an app: How does onboarding work? What's the time-to-value? What does the empty state look like? How does the first action feel? If it's a landing page: How's the scroll flow? Is there a logical progression from curiosity to action?
+ - **Page critique:** one public page or screenshot.
+ - **Marketing-site teardown:** positioning, proof, pricing path, and conversion.
+ - **Onboarding/activation teardown:** first-use journey and time to value.
+ - **Pricing/packaging teardown:** plans, value metric, comparison, and purchase friction.
+ - **Full-funnel teardown:** acquisition through activation and visible retention loops.
+ - **Comparison:** evaluate multiple products on the same scoped lenses.
- **7. Design & Craft**
- Visual hierarchy, whitespace, typography, color. Does the design support the message or distract from it? Is it distinctive or template-generic? Does it feel like someone cared?
+ Infer the target product, application context, and learning objective from the request. Ask one blocking question only when the product or intended application is genuinely missing.
- ## Phase 3: The Three Lists
+ ## Evidence protocol
- After analysis, distill into three lists:
+ 1. Record product, URLs or artifact names, capture date, viewport/device, authentication state, locale, and pages/states covered.
+ 2. Distinguish visible evidence, user-provided behavior, sourced fact, and inference.
+ 3. Cite URLs and describe the exact screen, state, or copy supporting each important finding. State access gaps rather than extrapolating.
+ 4. Do not infer conversion, retention, revenue, strategy, or user sentiment from UI alone.
- ```
- TEARDOWN — [Product Name]
- ════════════════════════════════════
+ ## Lenses
- WHAT'S WORKING:
- 1. [specific thing] — [why it works]
- 2. [specific thing] — [why it works]
- 3. [specific thing] — [why it works]
- (3-7 items)
+ Select only relevant lenses and explain material omissions:
- WHAT'S NOT:
- 1. [specific thing] — [what's wrong and why]
- 2. [specific thing] — [what's wrong and why]
- 3. [specific thing] — [what's wrong and why]
- (3-7 items)
+ - five-second comprehension and category;
+ - audience, alternative, differentiation, and message hierarchy;
+ - copy, information scent, calls to action, and objection handling;
+ - proof, trust, risk, privacy, and credibility;
+ - pricing, packaging, value metric, and purchase friction;
+ - navigation, primary task, empty/loading/error/recovery states;
+ - onboarding, activation, time to value, and progressive disclosure;
+ - accessibility, responsive behavior, performance cues, and inclusive design;
+ - visual hierarchy, typography, motion, and craft;
+ - visible sharing, collaboration, habit, or retention loops.
- STEAL THIS:
- 1. [specific tactic] — [how to apply it to YOUR product]
- 2. [specific tactic] — [how to apply it to YOUR product]
- 3. [specific tactic] — [how to apply it to YOUR product]
- (3-5 items. Each must reference the user's own project.)
+ ## Workflow
- ════════════════════════════════════
- ```
+ 1. Define scope and capture protocol.
+ 2. Walk the actual journey in order, recording evidence before evaluation.
+ 3. Evaluate relevant lenses against the product's audience and job, not generic taste.
+ 4. Rank findings by user impact, confidence, frequency/reach, and effort.
+ 5. Produce transferable principles with fit conditions, risks, and a concrete application to the user's product.
+ 6. If implementation is requested, inspect the target repository before proposing or making changes.
- ## Phase 4: Go Deeper (Optional)
+ ## Output contract
- If the user wants more, offer:
+ Return:
- - **Copy rewrite** — take their weakest section and rewrite it using the best tactic from the teardown
- - **Side-by-side** — compare two competitors on the same 7 lenses
- - **Apply it** — take the "Steal This" list and make the changes in the user's codebase
+ - scope and evidence ledger;
+ - concise journey or page narrative;
+ - findings table with evidence, lens, impact, confidence, and recommendation;
+ - strengths worth preserving;
+ - weaknesses or unanswered objections;
+ - transferable principles mapped to the user's product, including when not to apply them;
+ - access gaps and next evidence to collect.
## Verify
- ```
- [ ] Target product is clearly identified and fully read
- [ ] All 7 lenses analyzed (or explicitly skipped with reason)
- [ ] "What's Working" has specific observations, not generic praise
- [ ] "What's Not" has specific problems, not vague criticism
- [ ] "Steal This" items reference the user's own project — not generic advice
- [ ] Teardown is opinionated — takes a position, doesn't sit on the fence
- [ ] No "they could consider" or "it might be worth exploring" hedging
- [ ] Analysis is grounded in what's visible, not assumptions about their metrics
- ```
+ - Capture date, routes/states, viewport, and access boundary are explicit.
+ - Every important claim is tied to visible or sourced evidence.
+ - Pricing, onboarding, and product behavior are not inferred from a landing page unless marked unknown.
+ - Recommendations fit the target user's product and constraints.
+ - Transferable principles avoid copying expression or distinctive assets.
+ - Findings include impact, confidence, and limitations rather than unsupported certainty.