hvac-marketing-skills/skills/form-cro/evals/evals.json
Corey Haines 11e9ea811f feat: add evals for all 29 remaining skills (197 total evals across 32 skills)
Each skill now has 5-8 evals covering:
- Core framework usage with realistic prompts
- Casual trigger phrase variants
- Sub-type and section-specific coverage
- Boundary tests (skill deferral to related skills)
- Structured assertions for grading

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-04 13:37:01 -08:00

90 lines
6.3 KiB
JSON

{
"skill_name": "form-cro",
"evals": [
{
"id": 1,
"prompt": "Audit our demo request form. It currently has these fields: First Name, Last Name, Work Email, Phone Number, Company Name, Company Size, Job Title, Industry, Current Solution, Budget Range, and a 'Tell us about your needs' textarea. Our conversion rate is 3.1% and we want to improve it.",
"expected_output": "Should check for product-marketing-context.md first. Should apply the 'every field has a cost' principle — 11 fields is excessive for a demo form. Should reference the field cost data (3 fields baseline, 7+ fields = 25-50% conversion reduction). Should evaluate each field: which are essential for demo prep, which can be collected later or inferred. Should recommend cutting to essential fields (likely Work Email, Company Name, and maybe one qualifier). Should provide audit findings in the structured format (Issue, Impact, Fix, Priority). Should recommend Quick Wins, High-Impact Changes, and Test Ideas.",
"assertions": [
"Checks for product-marketing-context.md",
"Applies 'every field has a cost' principle",
"References field count impact data",
"Evaluates each field for necessity",
"Recommends cutting to essential fields",
"Provides findings in structured format (Issue, Impact, Fix, Priority)",
"Includes Quick Wins, High-Impact Changes, Test Ideas"
],
"files": []
},
{
"id": 2,
"prompt": "Our contact form just has Name, Email, and Message fields but we're getting a lot of spam submissions and low-quality leads. How do we fix this without adding too much friction?",
"expected_output": "Should apply the contact form type guidance. Should address spam with non-friction solutions first: honeypot fields, reCAPTCHA, server-side validation. Should then address lead quality: suggest adding one qualifying field (company name or budget range) to filter without excessive friction. Should apply the error handling guidance for validation. Should recommend form layout and submit button optimization. Should balance quality vs quantity in recommendations.",
"assertions": [
"Applies contact form type guidance",
"Recommends anti-spam solutions (honeypot, reCAPTCHA)",
"Suggests minimal qualifying fields for lead quality",
"Balances quality vs quantity",
"Addresses error handling and validation",
"Recommends non-friction solutions first"
],
"files": []
},
{
"id": 3,
"prompt": "should we use a single-step or multi-step form for our quote request? we need company info, project details, timeline, and budget.",
"expected_output": "Should trigger on casual phrasing. Should apply the multi-step form guidance — with this many required data types, multi-step is likely better. Should reference the threshold: multi-step recommended when more than 5-6 fields. Should recommend grouping by type (contact info → project details → budget/timeline). Should include progress indicator recommendation. Should apply best practices: easy questions first, save progress, allow back navigation. Should note that multi-step often increases completion for longer forms.",
"assertions": [
"Triggers on casual phrasing",
"Recommends multi-step based on field count",
"References the 5-6 field threshold for multi-step",
"Suggests logical field grouping",
"Recommends progress indicator",
"Applies multi-step best practices",
"Notes multi-step increases completion for longer forms"
],
"files": []
},
{
"id": 4,
"prompt": "How should we handle form errors? Users keep getting frustrated and abandoning our lead capture form when they hit validation errors.",
"expected_output": "Should apply the error handling guidance. Should recommend inline validation (not just on submit). Should provide specific error message examples (helpful, not generic). Should recommend: don't clear the form on error, focus on the problem field, show requirements upfront not after failure. Should address common validation UX issues: email format, phone format, required field indicators. Should provide examples of good vs bad error messages.",
"assertions": [
"Applies error handling guidance",
"Recommends inline validation",
"Provides specific error message examples",
"Recommends not clearing form on error",
"Recommends showing requirements upfront",
"Provides good vs bad error message examples",
"Addresses common validation UX issues"
],
"files": []
},
{
"id": 5,
"prompt": "We need to optimize our form for mobile. Over 60% of our traffic is mobile but our form conversion rate on mobile is half of desktop.",
"expected_output": "Should apply the mobile optimization guidance. Should recommend: larger touch targets (44px+ height), appropriate keyboard types (email, tel), autofill support, single column layout, sticky CTA button, reduce typing (use dropdowns, toggles). Should address mobile-specific form issues: viewport, font size, button placement, scroll behavior. Should recommend testing with actual devices.",
"assertions": [
"Applies mobile optimization guidance",
"Recommends larger touch targets (44px+)",
"Recommends appropriate keyboard types",
"Recommends autofill support",
"Recommends single column layout",
"Addresses mobile-specific issues",
"Recommends testing with actual devices"
],
"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/registration form optimization, not general form CRO. Should defer to or cross-reference the signup-flow-cro skill, which specifically handles signup, registration, and account creation flows. May provide general form friction advice but should make clear that signup-flow-cro is the right skill for signup forms.",
"assertions": [
"Recognizes this as signup flow optimization",
"References or defers to signup-flow-cro skill",
"Does not attempt full signup form optimization using general form CRO patterns"
],
"files": []
}
]
}