ai-product-audit
Skill1 starsChanged 6 months agoMITView on GitHub
Audit AI-generated products for product-pattern fit, scenario clarity, conversion readiness, and business outcome alignment. Use when the user wants to know whether a page prepares the user for the right scenario, moves them toward a concrete outcome, and converts inspiration into action.
In this repository11 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .cursor/rules/audit-live-workbench.mdcCursor rule
- accept-fiveSkill
- agent-diagnoseSkill
- auditSkill
- deploy-checkSkill
- flow-testSkill
- harnessSkill
- physical-flow-testSkill
- skill-studySkill
- visual-qaSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 7 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Use this skill to diagnose whether an AI-built product follows proven product patterns and converts inspiration into action.
- Distinctive: line 26
Before declaring a page "conversion-ready" or "shippable."
- Distinctive: line 32
Every finding needs three things: expected pattern, observed gap, and business risk.
- Distinctive: line 33
Use S0-S4 severity mapped to delivery and conversion risk, not subjective preference.
- Distinctive: line 37
Never claim a product "understands its user" without citing a specific page element and its failure.
- Very specific: line 42
Use references/product-pattern-rubric.md for the full dimension list.
- Very specific: line 43
Use references/category-pattern-catalog.md for category-specific pattern expectations.
- Distinctive: line 49
Scenario audit: ask the four Viba-inspired questions for each key page:
- Distinctive: line 66
Use the shared output shape for every finding.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI Product Audit
Use this skill to diagnose whether an AI-built product follows proven product patterns and converts inspiration into action.
product category + scenario -> pattern matching -> scenario audit -> conversion surface audit -> business reality check -> issue cards -> regression check -> lessonsThis skill must not judge a product by visual polish alone. It asks whether the page prepares the user for a believable next step and provides a path to reach it.
When To Use
- The target is a lifestyle, service, commerce, creator, SaaS, portfolio, directory, or dashboard product.
/auditor/visual-qaflags a pattern mismatch or vague value proposition.- The user wants to know whether the page converts inspiration into action.
- Batch-auditing multiple sites for product-pattern fitness.
- Before declaring a page "conversion-ready" or "shippable."
Core Rules
- Separate product-pattern evidence from visual evidence — a polished page can still have a broken scenario.
- Compare against the proven pattern for the product category, not personal taste.
- Every finding needs three things: expected pattern, observed gap, and business risk.
- Use S0-S4 severity mapped to delivery and conversion risk, not subjective preference.
- Preserve the shared output shape: Scope, Evidence, Findings, Severity, Reproduction, Fix Suggestion, Regression Check, Lessons.
- For batch audits, emit a summary table first, then progressive per-site details.
- Mark payment, irreversible submission, and production mutation as
SKIPPED-SAFEunless explicitly allowed. - Never claim a product "understands its user" without citing a specific page element and its failure.
Workflow
- Intake and scope: identify product category, intended scenario, business outcome, conversion surfaces, and audit depth.
- Use
references/product-pattern-rubric.mdfor the full dimension list. - Use
references/category-pattern-catalog.mdfor category-specific pattern expectations.
- Use
- Surface and pattern check: discover the visible page surface; compare each page against its category pattern.
- Apply the permission model before any click, form fill, or authenticated action.
- Mark pages or flows that cannot be safely tested as
SKIPPED-SAFE.
- Scenario audit: ask the four Viba-inspired questions for each key page:
- What scenario is this page preparing the user for?
- What self-image, business outcome, or action does it help the user move toward?
- Can the user see themselves in the next step?
- Is the page only inspiration, or does it convert inspiration into action?
- Conversion surface audit: for each identified CTA, form, booking flow, checkout, or signup path:
- Is the primary CTA specific and actionable?
- Does the page contain a working conversion surface (not just a brochure)?
- Is there a visible path from inspiration to action in fewer than 3 clicks?
- Business reality check: distinguish real products from templates.
- Is there operational depth (backend, database, CMS, auth, content system)?
- Is there a monetization path or demonstrated usage?
- Does the evidence (source, live, or physical) support a real business claim?
- Evidence assembly and output: produce issue cards, pattern-fit table, and copyable fix prompts.
- Use the shared output shape for every finding.
- Include a Pattern Fit table and a Scenario Audit table.
- Bundle fix prompts so the user can copy them directly into Claude Code, Lovable, v0, or Bol.
- Regression and lessons: convert repeated pattern failures into guardrail updates or benchmark labels.
- Propose updates to
CLAUDE.mdonly when the pattern appears in 3+ audits with clear evidence. - Append lessons to the audit ledger in
validation/for future five-pass reviews.
- Propose updates to
References
references/product-pattern-rubric.mdreferences/category-pattern-catalog.md../audit/references/progressive-reporting.md(for batch audits and multi-step runs)../visual-qa/references/aesthetic-quality-audit.md(for pattern reference and AI slop signals)../audit/references/permission-model.md(before any live or authenticated action)
Output Format
AI Product-Pattern Audit Summary
Target:
Product Category:
Intended Scenario:
Pattern Fit Score:
Main Business Risk:
Fix First:
Pattern Fit Table
| Expected Pattern | Observed | Gap | Risk | S0-S4 |
| --- | --- | --- | --- | --- |
Scenario Audit
| Question | Answer | Evidence | Risk |
| --- | --- | --- | --- |
Conversion Surface Map
| Surface | Present | Actionable | Evidence |
| --- | --- | --- | --- |
Business Reality
| Signal | Present | Evidence |
| --- | --- | --- |
Issue Cards
<S0-S4> - <Product Pattern Issue>
- Area:
- URL:
- Live position:
- Expected pattern:
- Observed:
- Business risk:
- Fix:
- Copy prompt:
- Regression check:
Copyable Fix Pack
1. <ready-to-copy prompt>
2. <ready-to-copy prompt>
3. <ready-to-copy prompt>
LessonsAnti-Patterns
- Judging product quality by visual polish alone — visual QA and product-pattern audit are different dimensions.
- Applying SaaS patterns to a local service site, or portfolio patterns to a commerce site.
- Treating "vibe" or "mood" as a substitute for scenario clarity.
- Missing the "next step" test — if the user cannot describe what happens after clicking, the scenario is broken.
- Batch-auditing without first categorizing each site — mixed-category batches produce misleading summaries.
- Claiming a page "converts" because it has a CTA — the CTA must be specific, actionable, and lead to a working next step.
- Using product-pattern findings to rewrite copy subjectively — always tie the fix to a pattern mismatch, not a taste preference.
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
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.

