fix: improve evals with assertions, coverage gaps, and boundary tests

Changes per skill:

page-cro (5 -> 7 evals):
- Added structured assertions array to all evals
- Added Blog Post CRO eval (was missing)
- Added boundary eval: signup form should defer to signup-flow-cro
- Replaced feature page A/B test eval with fuller feature page eval

copywriting (5 -> 7 evals):
- Added structured assertions array to all evals
- Added About Page eval with voice/tone adaptation test
- Added boundary eval: email sequence should defer to email-sequence
- Added Quick Quality Check eval (buzzword/jargon/exclamation detection)
- Added meta content assertion to homepage eval

seo-audit (5 -> 8 evals):
- Added structured assertions array to all evals
- Added Local Business eval (NAP, GBP, location pages)
- Added Core Web Vitals / site speed eval
- Added boundary eval: FAQ schema should defer to schema-markup

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

skills/copywriting/evals/evals.json

@@ -4,31 +4,107 @@
     {
       "id": 1,
       "prompt": "Write homepage copy for a SaaS tool that automates employee onboarding. Target audience is HR directors at mid-size companies (200-2000 employees). Main differentiator is that it integrates with all major HRIS systems and cuts onboarding time from 2 weeks to 2 days.",
-      "expected_output": "Should check for product-marketing-context.md first. Should write full page copy organized by section: Headline, Subheadline, CTA (above the fold), then Social Proof, Problem/Pain, Solution/Benefits, How It Works, Objection Handling, and Final CTA. Should follow copywriting principles: clarity over cleverness, benefits over features, specificity (use the '2 weeks to 2 days' stat), customer language. Headline should communicate core value proposition. CTAs should be action-oriented ('Start Free Trial' not 'Submit'). Should provide 2-3 headline alternatives with rationale. Should include annotations explaining key copy choices.",
+      "expected_output": "Should check for product-marketing-context.md first. Should write full page copy organized by section: Headline, Subheadline, CTA (above the fold), then Social Proof, Problem/Pain, Solution/Benefits, How It Works, Objection Handling, and Final CTA. Should follow copywriting principles: clarity over cleverness, benefits over features, specificity (use the '2 weeks to 2 days' stat), customer language. Headline should communicate core value proposition. CTAs should be action-oriented ('Start Free Trial' not 'Submit'). Should provide 2-3 headline alternatives with rationale. Should include annotations explaining key copy choices. Should include meta content (SEO page title and meta description).",
+      "assertions": [
+        "Checks for product-marketing-context.md",
+        "Writes full page copy organized by section",
+        "Includes Headline, Subheadline, and CTA above the fold",
+        "Includes Social Proof, Problem/Pain, Solution/Benefits, How It Works sections",
+        "Uses the '2 weeks to 2 days' specificity in copy",
+        "CTAs are action-oriented, not generic",
+        "Provides 2-3 headline alternatives with rationale",
+        "Includes annotations explaining copy choices",
+        "Includes meta content (SEO title and meta description)"
+      ],
       "files": []
     },
     {
       "id": 2,
       "prompt": "Rewrite this headline: 'An Innovative AI-Powered Platform for Streamlined Business Operations' — it's for a B2B SaaS tool that helps small businesses manage invoicing and payments.",
       "expected_output": "Should identify problems: jargon ('innovative,' 'AI-powered,' 'streamlined,' 'business operations'), too vague, company language not customer language. Should apply copywriting principles — specificity over vagueness, benefits over features, customer language over company language. Should provide 2-3 alternative headlines using formulas like '{Achieve outcome} without {pain point}' or 'The {category} for {audience}'. Each alternative should include rationale. Should also suggest a subheadline that adds specificity.",
+      "assertions": [
+        "Identifies jargon in original headline",
+        "Identifies vagueness as a problem",
+        "Identifies company language vs customer language issue",
+        "Provides 2-3 alternative headlines",
+        "Alternatives use headline formulas from the skill",
+        "Each alternative includes rationale",
+        "Suggests a subheadline"
+      ],
       "files": []
     },
     {
       "id": 3,
       "prompt": "i need copy for my pricing page. we have three plans: starter ($29/mo), pro ($79/mo), business ($199/mo). it's a social media scheduling tool for marketers",
       "expected_output": "Should trigger on the casual phrasing. Should ask or infer audience context. Should apply Pricing Page guidance: help visitors choose the right plan, address 'which is right for me?' anxiety, make recommended plan obvious. Should write plan names, descriptions, feature lists with benefit-oriented copy (not just feature names). Should include a page headline that addresses the pricing decision. CTAs should be specific per plan. Should handle objection handling (FAQ copy). Should provide alternatives for key elements.",
+      "assertions": [
+        "Triggers on casual phrasing",
+        "Applies Pricing Page guidance",
+        "Addresses 'which plan is right for me' anxiety",
+        "Makes recommended plan obvious",
+        "Writes benefit-oriented feature copy, not just feature names",
+        "Includes page headline",
+        "CTAs are specific per plan",
+        "Includes FAQ or objection handling copy",
+        "Provides alternatives for key elements"
+      ],
       "files": []
     },
     {
       "id": 4,
-      "prompt": "Write a landing page for a free ebook download: '10 Ways to Reduce Customer Churn.' The target audience is Customer Success managers at SaaS companies.",
+      "prompt": "Write copy for our About page. We're a 3-person startup that built a developer tool for database migrations. Founded because we kept losing data during migrations at our last jobs. Tone should be professional but human.",
-      "expected_output": "Should apply Landing Page guidance: single message, single CTA, complete argument on one page. Should write a compelling headline that communicates the value of the ebook (not just its title). Should address the reader's pain (churn), build credibility for the content, and minimize friction to download. CTA should be specific ('Download the Free Guide' not 'Submit'). Should keep form fields minimal (email + name at most). Copy should follow all style rules: simple over complex, specific over vague, active voice, confident tone. Should provide headline alternatives.",
+      "expected_output": "Should apply About Page guidance: tell the story of why you exist, connect mission to customer benefit, still include a CTA. Should adapt voice and tone to 'professional but human' as specified. Should tell the founder origin story authentically. Should connect the personal pain to the customer's pain. Should include a CTA even on the About page. Copy should follow style rules: active voice, confident, specific. Should NOT be overly corporate or generic.",
+      "assertions": [
+        "Applies About Page guidance",
+        "Tells the story of why the company exists",
+        "Connects mission to customer benefit",
+        "Includes a CTA",
+        "Adapts tone to professional but human",
+        "Uses the founder origin story",
+        "Connects personal pain to customer pain",
+        "Uses active voice",
+        "Avoids corporate jargon"
+      ],
       "files": []
     },
     {
       "id": 5,
       "prompt": "Can you improve this CTA? We currently have 'Learn More' on our feature page for our analytics dashboard product.",
       "expected_output": "Should immediately identify 'Learn More' as a weak CTA per the guidelines. Should apply the CTA formula: [Action Verb] + [What They Get] + [Qualifier]. Should provide 2-3 strong alternatives like 'See the Dashboard in Action,' 'Start Your Free Trial,' or 'Explore Analytics Features.' Each alternative should include rationale and context for when it works best. Should also consider CTA hierarchy — whether this is a primary or secondary CTA, and suggest complementary CTAs if relevant.",
+      "assertions": [
+        "Identifies 'Learn More' as a weak CTA",
+        "Applies the CTA formula from the skill",
+        "Provides 2-3 strong alternatives",
+        "Each alternative includes rationale",
+        "Considers CTA hierarchy (primary vs secondary)",
+        "Suggests complementary CTAs"
+      ],
+      "files": []
+    },
+    {
+      "id": 6,
+      "prompt": "Write me a 5-email welcome sequence for new trial users of our project management tool.",
+      "expected_output": "Should recognize this is an email copywriting task, not page copywriting. Should defer to or cross-reference the email-sequence skill, which specifically handles email sequences, drip campaigns, and lifecycle emails. May provide brief general guidance but should make clear that email-sequence is the right skill for this task.",
+      "assertions": [
+        "Recognizes this as email sequence work",
+        "References or defers to email-sequence skill",
+        "Does not attempt to write a full email sequence using page copywriting patterns"
+      ],
+      "files": []
+    },
+    {
+      "id": 7,
+      "prompt": "Review this copy and tell me what's wrong: 'We are extremely excited to announce our revolutionary, cutting-edge platform that will totally transform how businesses optimize their workflows! Sign up now!!'",
+      "expected_output": "Should apply the Quick Quality Check. Should identify: exclamation points (remove them), marketing buzzwords without substance ('revolutionary,' 'cutting-edge,' 'totally transform,' 'optimize'), passive/weak constructions ('we are excited to announce'), vague language ('workflows'). Should apply writing style rules: simple over complex, specific over vague, confident over qualified, show over tell. Should rewrite the copy following these principles. Should provide 2-3 alternatives.",
+      "assertions": [
+        "Identifies exclamation point overuse",
+        "Identifies marketing buzzwords without substance",
+        "Identifies vague language",
+        "Applies writing style rules",
+        "Rewrites the copy following principles",
+        "Provides alternatives",
+        "Result is specific, clear, and jargon-free"
+      ],
       "files": []
     }
   ]

skills/page-cro/evals/evals.json

@@ -5,30 +5,106 @@
       "id": 1,
       "prompt": "Here's my SaaS landing page: https://example.com/product. We get about 5,000 visitors/month from Google Ads but only 1.2% convert to free trial signups. Can you help me figure out what's wrong?",
       "expected_output": "Should check for product-marketing-context.md first. Should identify page type (landing page) and conversion goal (free trial signup). Should analyze across the CRO framework dimensions: value proposition clarity, headline effectiveness, CTA placement/copy/hierarchy, visual hierarchy, trust signals, objection handling, and friction points. Should provide recommendations organized as Quick Wins, High-Impact Changes, and Test Ideas. Should note the message match issue between Google Ads and landing page. Should provide 2-3 headline and CTA copy alternatives with rationale.",
+      "assertions": [
+        "Checks for product-marketing-context.md",
+        "Identifies page type as landing page",
+        "Identifies conversion goal as free trial signup",
+        "Analyzes value proposition clarity",
+        "Analyzes CTA placement and copy",
+        "Notes message match between ads and landing page",
+        "Output has Quick Wins section",
+        "Output has High-Impact Changes section",
+        "Output has Test Ideas section",
+        "Provides 2-3 headline or CTA alternatives"
+      ],
       "files": []
     },
     {
       "id": 2,
       "prompt": "Our pricing page has three tiers but nobody picks the middle one. 60% choose the cheapest plan and 30% bounce entirely. What should we change?",
       "expected_output": "Should apply the Pricing Page CRO framework. Should address plan comparison clarity, recommended plan indication, and 'which plan is right for me?' anxiety. Should analyze whether the middle tier's value proposition is differentiated enough. Should recommend trust signals and social proof near pricing. Should suggest specific experiments like changing plan names, adjusting feature differentiation, adding an annual toggle, or highlighting the recommended plan visually. Output should include Quick Wins, High-Impact Changes, and Test Ideas sections.",
+      "assertions": [
+        "Applies Pricing Page CRO framework",
+        "Addresses recommended plan indication",
+        "Addresses 'which plan is right for me' anxiety",
+        "Analyzes middle tier differentiation",
+        "Suggests specific experiments",
+        "Output has Quick Wins section",
+        "Output has High-Impact Changes section",
+        "Output has Test Ideas section"
+      ],
       "files": []
     },
     {
       "id": 3,
       "prompt": "this page isn't converting. can you take a look? it's our homepage for a B2B project management tool",
       "expected_output": "Should trigger on the casual 'this page isn't converting' phrasing. Should identify this as a Homepage CRO analysis. Should ask clarifying questions about current conversion rate, traffic sources, and conversion goal. Should apply the full CRO Analysis Framework starting with value proposition clarity. Should address the homepage-specific guidance: serving multiple audiences, leading with broadest value prop, and providing clear paths for different visitor intents. Should provide structured output with Quick Wins, High-Impact Changes, Test Ideas, and Copy Alternatives.",
+      "assertions": [
+        "Triggers on casual phrasing",
+        "Identifies as Homepage CRO",
+        "Asks about current conversion rate",
+        "Asks about traffic sources",
+        "Applies CRO Analysis Framework",
+        "Addresses serving multiple audiences",
+        "Addresses clear paths for different visitor intents",
+        "Output has structured sections"
+      ],
       "files": []
     },
     {
       "id": 4,
-      "prompt": "I want to A/B test some changes on my feature page. What should I test first?",
+      "prompt": "We have a blog that gets 20k organic visits/month but almost nobody clicks through to our product. How do we get more conversions from blog readers?",
-      "expected_output": "Should apply the Feature Page CRO framework (connect feature to benefit, use cases, clear path to try/buy). Should reference the experiments section and suggest prioritized test ideas for hero section, trust signals, and CTA variations. Should recommend testing one variable at a time. Should cross-reference ab-test-setup skill for proper test implementation. Output should include specific, actionable hypotheses — not vague suggestions.",
+      "expected_output": "Should apply the Blog Post CRO framework. Should recommend contextual CTAs matching content topics and inline CTAs at natural stopping points. Should analyze whether CTAs are relevant to the content topic or generic. Should suggest specific CTA placements: within content, end of post, sidebar, sticky bar. Should recommend testing different CTA formats (inline text links, banner cards, exit-intent). Should cross-reference copywriting skill for CTA copy improvement.",
+      "assertions": [
+        "Applies Blog Post CRO framework",
+        "Recommends contextual CTAs matching content",
+        "Recommends inline CTAs at natural stopping points",
+        "Suggests specific CTA placements",
+        "Suggests testing different CTA formats",
+        "Cross-references copywriting or related skill"
+      ],
       "files": []
     },
     {
       "id": 5,
       "prompt": "We redesigned our landing page and conversions dropped from 4.2% to 2.8%. Here's the new page. What went wrong?",
       "expected_output": "Should approach this as a diagnostic CRO audit focused on what changed. Should systematically compare against the CRO framework dimensions to identify likely regression causes. Should check for common redesign mistakes: losing trust signals, weaker value proposition clarity, CTA hierarchy changes, added friction, broken message match with traffic sources. Should provide specific fixes organized by likely impact. Should recommend reverting high-risk changes while testing others.",
+      "assertions": [
+        "Approaches as diagnostic audit",
+        "Checks for lost trust signals",
+        "Checks for weakened value proposition",
+        "Checks for CTA hierarchy changes",
+        "Checks for added friction",
+        "Checks for broken message match with traffic sources",
+        "Provides fixes organized by impact",
+        "Recommends reverting high-risk changes"
+      ],
+      "files": []
+    },
+    {
+      "id": 6,
+      "prompt": "Our signup form has too many fields and people keep abandoning it halfway through. Can you help optimize it?",
+      "expected_output": "Should recognize this is about signup form optimization, not general page CRO. Should defer to or cross-reference the signup-flow-cro skill, which specifically handles signup, registration, and account creation flows. May provide some general friction reduction advice but should make clear that signup-flow-cro is the right skill for this task.",
+      "assertions": [
+        "Recognizes this as signup flow optimization",
+        "References or defers to signup-flow-cro skill",
+        "Does not attempt full page-cro analysis on a form"
+      ],
+      "files": []
+    },
+    {
+      "id": 7,
+      "prompt": "Review this feature page for our API monitoring tool. Most traffic comes from organic search for 'API monitoring tools'. We want them to start a free trial.",
+      "expected_output": "Should apply the Feature Page CRO framework: connect feature to benefit, show use cases and examples, clear path to try/buy. Should reference the experiments section and suggest prioritized test ideas for hero section, trust signals, and CTA variations. Should note the organic search traffic source and check for message match with search intent. Should cross-reference ab-test-setup skill for proper test implementation.",
+      "assertions": [
+        "Applies Feature Page CRO framework",
+        "Connects features to benefits",
+        "Suggests use cases and examples",
+        "Provides clear path to try/buy",
+        "Notes organic traffic source and search intent match",
+        "Suggests specific experiment hypotheses",
+        "Cross-references ab-test-setup skill"
+      ],
       "files": []
     }
   ]

skills/seo-audit/evals/evals.json

@@ -5,30 +5,131 @@
       "id": 1,
       "prompt": "Can you do an SEO audit of our SaaS website? We're getting about 2,000 organic visits/month but feel like we should be getting more. URL: https://example.com",
       "expected_output": "Should check for product-marketing-context.md first. Should ask clarifying questions about priority keywords, Search Console access, recent changes, and competitors. Should follow the audit framework priority order: Crawlability & Indexation, Technical Foundations, On-Page Optimization, Content Quality, Authority & Links. Should check robots.txt, XML sitemap, site architecture. Should evaluate title tags, meta descriptions, heading structure, and content optimization. Should NOT report on schema markup based solely on web_fetch (must note the detection limitation). Output should follow the Audit Report Structure: Executive Summary, Technical SEO Findings, On-Page SEO Findings, Content Findings, and Prioritized Action Plan.",
+      "assertions": [
+        "Checks for product-marketing-context.md",
+        "Asks clarifying questions about keywords, Search Console, recent changes",
+        "Follows audit priority order: crawlability first, then technical, on-page, content, authority",
+        "Checks robots.txt and XML sitemap",
+        "Evaluates title tags, meta descriptions, heading structure",
+        "Does NOT claim 'no schema found' based on web_fetch alone",
+        "Notes schema markup detection limitation",
+        "Output has Executive Summary",
+        "Output has Prioritized Action Plan",
+        "Each finding has Issue, Impact, Evidence, Fix, and Priority"
+      ],
       "files": []
     },
     {
       "id": 2,
       "prompt": "Why am I not ranking for 'project management software'? We have a page targeting that keyword but it's stuck on page 3.",
       "expected_output": "Should trigger on the casual 'why am I not ranking' phrasing. Should investigate both on-page and off-page factors. On-page: check title tag, H1, URL alignment with keyword; evaluate content depth vs competitors; check for keyword cannibalization. Technical: check indexation status, canonical tags, crawlability. Content quality: assess E-E-A-T signals, content depth, user engagement. Should provide specific, actionable fixes organized by priority. Should mention competitive analysis against current top-ranking pages.",
+      "assertions": [
+        "Triggers on casual 'why am I not ranking' phrasing",
+        "Checks title tag, H1, URL alignment with target keyword",
+        "Evaluates content depth vs competitors",
+        "Checks for keyword cannibalization",
+        "Checks indexation status and canonical tags",
+        "Assesses E-E-A-T signals",
+        "Mentions competitive analysis against top-ranking pages",
+        "Provides actionable fixes organized by priority"
+      ],
       "files": []
     },
     {
       "id": 3,
       "prompt": "We just migrated from WordPress to Next.js and our organic traffic dropped 40% in the last month. Help!",
       "expected_output": "Should treat this as an urgent migration diagnostic. Should immediately check: redirect mapping (301s from old URLs to new), canonical tags on new pages, robots.txt not blocking crawlers, XML sitemap submitted and updated, meta tags preserved. Should check for common migration issues: redirect chains/loops, soft 404s, lost internal links, changed URL structures without redirects. Should reference Search Console coverage report for indexation issues. Should provide a prioritized recovery plan with critical fixes first. Should mention monitoring timeline expectations (recovery can take weeks).",
+      "assertions": [
+        "Treats as urgent migration diagnostic",
+        "Checks redirect mapping (301s)",
+        "Checks canonical tags on new pages",
+        "Checks robots.txt not blocking crawlers",
+        "Checks XML sitemap updated and submitted",
+        "Checks for redirect chains or loops",
+        "Checks for soft 404s",
+        "References Search Console coverage report",
+        "Provides prioritized recovery plan",
+        "Mentions recovery timeline expectations"
+      ],
       "files": []
     },
     {
       "id": 4,
       "prompt": "Review the technical SEO of our e-commerce site. We have about 50,000 products and use faceted navigation.",
       "expected_output": "Should focus on e-commerce-specific technical issues: faceted navigation creating duplicate content, crawl budget management for large product catalog, parameterized URLs, product schema markup (with the caveat about detection limitations). Should check for thin category pages, duplicate product descriptions, out-of-stock page handling. Should address crawl budget issues: pagination, infinite scroll handling, session IDs in URLs. Should provide structured findings with Impact ratings and specific fixes.",
+      "assertions": [
+        "Addresses faceted navigation duplicate content",
+        "Addresses crawl budget for large catalog",
+        "Checks for parameterized URL issues",
+        "Mentions product schema with detection limitation caveat",
+        "Checks for thin category pages",
+        "Checks for duplicate product descriptions",
+        "Addresses out-of-stock page handling",
+        "Addresses pagination and infinite scroll",
+        "Findings include Impact ratings and specific fixes"
+      ],
       "files": []
     },
     {
       "id": 5,
       "prompt": "Can you check our blog posts for on-page SEO issues? We publish 4 posts per week but traffic has been flat for 6 months.",
       "expected_output": "Should apply the Content/Blog Sites framework: check for outdated content not refreshed, keyword cannibalization, missing topical clustering, poor internal linking, missing author pages. Should audit on-page elements: title tags, meta descriptions, heading structure, keyword targeting per post. Should assess E-E-A-T signals for blog content. Should check for content depth issues and whether posts answer search intent. Should recommend a content audit process and provide a prioritized action plan for the existing content library.",
+      "assertions": [
+        "Applies Content/Blog Sites framework",
+        "Checks for outdated content",
+        "Checks for keyword cannibalization",
+        "Checks for topical clustering",
+        "Checks for internal linking quality",
+        "Checks for author pages and E-E-A-T signals",
+        "Audits title tags, meta descriptions, heading structure",
+        "Assesses whether content answers search intent",
+        "Recommends content audit process",
+        "Provides prioritized action plan"
+      ],
+      "files": []
+    },
+    {
+      "id": 6,
+      "prompt": "I run a local plumbing business with 3 locations. My website barely shows up when people search for 'plumber near me' in our areas. What's wrong?",
+      "expected_output": "Should apply the Local Business site-type framework. Should check for: inconsistent NAP (Name, Address, Phone) across the site, missing local schema markup (with detection limitation caveat), Google Business Profile optimization, missing individual location pages for each of the 3 locations, and missing local content. Should also check standard technical and on-page factors. Should recommend local-specific fixes: location-specific pages with unique content, local schema on each, GBP optimization, citation consistency.",
+      "assertions": [
+        "Applies Local Business framework",
+        "Checks NAP consistency",
+        "Checks for local schema markup with detection caveat",
+        "Addresses Google Business Profile optimization",
+        "Recommends individual location pages for each location",
+        "Recommends local content strategy",
+        "Checks standard technical SEO factors too",
+        "Provides prioritized local SEO action plan"
+      ],
+      "files": []
+    },
+    {
+      "id": 7,
+      "prompt": "Our site loads really slowly, especially on mobile. Pages take 5-6 seconds to load. Is this hurting our SEO?",
+      "expected_output": "Should focus on Site Speed and Core Web Vitals. Should explain CWV thresholds: LCP < 2.5s, INP < 200ms, CLS < 0.1, and that 5-6s load time is well above acceptable. Should investigate speed factors: server response time (TTFB), image optimization, JavaScript execution, CSS delivery, caching headers, CDN usage, font loading. Should recommend specific tools: PageSpeed Insights, WebPageTest, Chrome DevTools, Search Console CWV report. Should explain that yes, page speed is a ranking factor and directly impacts SEO. Should provide prioritized fixes.",
+      "assertions": [
+        "Focuses on Core Web Vitals",
+        "Explains CWV thresholds (LCP, INP, CLS)",
+        "Identifies 5-6s as well above acceptable",
+        "Investigates specific speed factors",
+        "Recommends specific diagnostic tools",
+        "Confirms page speed impacts SEO rankings",
+        "Provides prioritized speed fixes",
+        "Addresses mobile-specific performance"
+      ],
+      "files": []
+    },
+    {
+      "id": 8,
+      "prompt": "I want to add FAQ schema to my product pages. Can you help me set that up?",
+      "expected_output": "Should recognize this is a schema markup implementation task, not an SEO audit. Should defer to or cross-reference the schema-markup skill, which specifically handles structured data implementation including FAQ schema. May briefly mention that FAQ schema can enable rich results, but should make clear that schema-markup is the right skill for implementation.",
+      "assertions": [
+        "Recognizes this as schema markup implementation",
+        "References or defers to schema-markup skill",
+        "Does not attempt a full SEO audit",
+        "May briefly mention FAQ schema benefits"
+      ],
       "files": []
     }
   ]